powerel9
/
glusterfs
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
2 Commits
1 Branch
0 Tags
8.3 MiB
Branch: master
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from 'master'
${ noResults }
glusterfs/SOURCES/0098-mgmt-shd-Implement-mul...

4618 lines
155 KiB
Raw Permalink Normal View History Unescape

initial package creation Signed-off-by: Toshaan Bharvani <toshaan@powerel.org>
3 years ago
From 2cede2b87fb3e3e0673be9cf67e7d6eec3f7879c Mon Sep 17 00:00:00 2001
From: Mohammed Rafi KC <rkavunga@redhat.com>
Date: Mon, 25 Feb 2019 10:05:32 +0530
Subject: [PATCH 098/124] mgmt/shd: Implement multiplexing in self heal daemon
Problem:
Shd daemon is per node, which means they create a graph
with all volumes on it. While this is a great for utilizing
resources, it is so good in terms of performance and managebility.
Because self-heal daemons doesn't have capability to automatically
reconfigure their graphs. So each time when any configurations
changes happens to the volumes(replicate/disperse), we need to restart
shd to bring the changes into the graph.
Because of this all on going heal for all other volumes has to be
stopped in the middle, and need to restart all over again.
Solution:
This changes makes shd as a per volume daemon, so that the graph
will be generated for each volumes.
When we want to start/reconfigure shd for a volume, we first search
for an existing shd running on the node, if there is none, we will
start a new process. If already a daemon is running for shd, then
we will simply detach a graph for a volume and reatach the updated
graph for the volume. This won't touch any of the on going operations
for any other volumes on the shd daemon.
Example of an shd graph when it is per volume
graph
-----------------------
| debug-iostat |
-----------------------
/ | \
/ | \
--------- --------- ----------
| AFR-1 | | AFR-2 | | AFR-3 |
-------- --------- ----------
A running shd daemon with 3 volumes will be like-->
graph
-----------------------
| debug-iostat |
-----------------------
/ | \
/ | \
------------ ------------ ------------
| volume-1 | | volume-2 | | volume-3 |
------------ ------------ ------------
Backport of: https://review.gluster.org/#/c/glusterfs/+/22075
>Change-Id: Idcb2698be3eeb95beaac47125565c93370afbd99
>fixes: bz#1659708
>Signed-off-by: Mohammed Rafi KC <rkavunga@redhat.com>
Change-Id: I38b6e2bb62edd818e460ccf6e9740364dc676876
BUG: 1471742
Signed-off-by: Mohammed Rafi KC <rkavunga@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/167830
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Atin Mukherjee <amukherj@redhat.com>
---
glusterfsd/src/glusterfsd-messages.h | 3 +-
glusterfsd/src/glusterfsd-mgmt.c | 238 ++++++-
glusterfsd/src/glusterfsd.c | 18 -
libglusterfs/src/defaults-tmpl.c | 19 +-
libglusterfs/src/glusterfs/glusterfs.h | 7 +
libglusterfs/src/glusterfs/libglusterfs-messages.h | 4 +-
libglusterfs/src/glusterfs/xlator.h | 3 +
libglusterfs/src/graph.c | 451 +++++++++++++
libglusterfs/src/graph.y | 3 +
libglusterfs/src/libglusterfs.sym | 5 +
libglusterfs/src/statedump.c | 3 +-
libglusterfs/src/xlator.c | 16 +
rpc/rpc-lib/src/protocol-common.h | 2 +
tests/basic/glusterd/heald.t | 49 +-
.../reset-brick-and-daemons-follow-quorum.t | 8 +-
tests/volume.rc | 6 +-
xlators/mgmt/glusterd/src/Makefile.am | 6 +-
xlators/mgmt/glusterd/src/glusterd-brick-ops.c | 2 +-
xlators/mgmt/glusterd/src/glusterd-conn-mgmt.c | 42 ++
xlators/mgmt/glusterd/src/glusterd-conn-mgmt.h | 4 +-
xlators/mgmt/glusterd/src/glusterd-gfproxyd-svc.c | 3 +-
xlators/mgmt/glusterd/src/glusterd-handler.c | 11 +-
xlators/mgmt/glusterd/src/glusterd-handshake.c | 21 +
xlators/mgmt/glusterd/src/glusterd-mem-types.h | 1 +
xlators/mgmt/glusterd/src/glusterd-messages.h | 4 +-
xlators/mgmt/glusterd/src/glusterd-op-sm.c | 84 ++-
.../mgmt/glusterd/src/glusterd-shd-svc-helper.c | 140 ++++
.../mgmt/glusterd/src/glusterd-shd-svc-helper.h | 45 ++
xlators/mgmt/glusterd/src/glusterd-shd-svc.c | 540 ++++++++++++++--
xlators/mgmt/glusterd/src/glusterd-shd-svc.h | 17 +-
xlators/mgmt/glusterd/src/glusterd-sm.c | 12 +-
xlators/mgmt/glusterd/src/glusterd-snapd-svc.c | 3 +-
xlators/mgmt/glusterd/src/glusterd-statedump.c | 3 -
xlators/mgmt/glusterd/src/glusterd-svc-helper.c | 715 ++++++++++++++++++++-
xlators/mgmt/glusterd/src/glusterd-svc-helper.h | 40 +-
xlators/mgmt/glusterd/src/glusterd-svc-mgmt.c | 246 +++++--
xlators/mgmt/glusterd/src/glusterd-svc-mgmt.h | 27 +
xlators/mgmt/glusterd/src/glusterd-tier.c | 3 +-
xlators/mgmt/glusterd/src/glusterd-tierd-svc.c | 4 +-
xlators/mgmt/glusterd/src/glusterd-utils.c | 137 +++-
xlators/mgmt/glusterd/src/glusterd-utils.h | 4 +
xlators/mgmt/glusterd/src/glusterd-volgen.c | 60 +-
xlators/mgmt/glusterd/src/glusterd-volgen.h | 11 +-
xlators/mgmt/glusterd/src/glusterd-volume-ops.c | 8 +-
xlators/mgmt/glusterd/src/glusterd.c | 12 +-
xlators/mgmt/glusterd/src/glusterd.h | 30 +-
xlators/protocol/client/src/client.c | 31 +-
47 files changed, 2810 insertions(+), 291 deletions(-)
create mode 100644 xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.c
create mode 100644 xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.h
diff --git a/glusterfsd/src/glusterfsd-messages.h b/glusterfsd/src/glusterfsd-messages.h
index 602cd9e..94312a5 100644
--- a/glusterfsd/src/glusterfsd-messages.h
+++ b/glusterfsd/src/glusterfsd-messages.h
@@ -35,6 +35,7 @@ GLFS_MSGID(GLUSTERFSD, glusterfsd_msg_1, glusterfsd_msg_2, glusterfsd_msg_3,
glusterfsd_msg_28, glusterfsd_msg_29, glusterfsd_msg_30,
glusterfsd_msg_31, glusterfsd_msg_32, glusterfsd_msg_33,
glusterfsd_msg_34, glusterfsd_msg_35, glusterfsd_msg_36,
- glusterfsd_msg_37, glusterfsd_msg_38, glusterfsd_msg_39);
+ glusterfsd_msg_37, glusterfsd_msg_38, glusterfsd_msg_39,
+ glusterfsd_msg_40, glusterfsd_msg_41, glusterfsd_msg_42);
#endif /* !_GLUSTERFSD_MESSAGES_H_ */
diff --git a/glusterfsd/src/glusterfsd-mgmt.c b/glusterfsd/src/glusterfsd-mgmt.c
index a6c3db5..a89c980 100644
--- a/glusterfsd/src/glusterfsd-mgmt.c
+++ b/glusterfsd/src/glusterfsd-mgmt.c
@@ -48,7 +48,20 @@ int
glusterfs_graph_unknown_options(glusterfs_graph_t *graph);
int
emancipate(glusterfs_ctx_t *ctx, int ret);
+int
+glusterfs_process_svc_attach_volfp(glusterfs_ctx_t *ctx, FILE *fp,
+ char *volfile_id, char *checksum);
+int
+glusterfs_mux_volfile_reconfigure(FILE *newvolfile_fp, glusterfs_ctx_t *ctx,
+ gf_volfile_t *volfile_obj, char *checksum);
+int
+glusterfs_process_svc_attach_volfp(glusterfs_ctx_t *ctx, FILE *fp,
+ char *volfile_id, char *checksum);
+int
+glusterfs_process_svc_detach(glusterfs_ctx_t *ctx, gf_volfile_t *volfile_obj);
+gf_boolean_t
+mgmt_is_multiplexed_daemon(char *name);
int
mgmt_cbk_spec(struct rpc_clnt *rpc, void *mydata, void *data)
{
@@ -62,6 +75,96 @@ mgmt_cbk_spec(struct rpc_clnt *rpc, void *mydata, void *data)
}
int
+mgmt_process_volfile(const char *volfile, ssize_t size, char *volfile_id)
+{
+ glusterfs_ctx_t *ctx = NULL;
+ int ret = 0;
+ FILE *tmpfp = NULL;
+ gf_volfile_t *volfile_obj = NULL;
+ gf_volfile_t *volfile_tmp = NULL;
+ char sha256_hash[SHA256_DIGEST_LENGTH] = {
+ 0,
+ };
+ int tmp_fd = -1;
+ char template[] = "/tmp/glfs.volfile.XXXXXX";
+
+ glusterfs_compute_sha256((const unsigned char *)volfile, size, sha256_hash);
+ ctx = THIS->ctx;
+ LOCK(&ctx->volfile_lock);
+ {
+ list_for_each_entry(volfile_obj, &ctx->volfile_list, volfile_list)
+ {
+ if (!strcmp(volfile_id, volfile_obj->vol_id)) {
+ if (!memcmp(sha256_hash, volfile_obj->volfile_checksum,
+ sizeof(volfile_obj->volfile_checksum))) {
+ UNLOCK(&ctx->volfile_lock);
+ gf_msg(THIS->name, GF_LOG_INFO, 0, glusterfsd_msg_40,
+ "No change in volfile, continuing");
+ goto out;
+ }
+ volfile_tmp = volfile_obj;
+ break;
+ }
+ }
+
+ /* coverity[secure_temp] mkstemp uses 0600 as the mode */
+ tmp_fd = mkstemp(template);
+ if (-1 == tmp_fd) {
+ UNLOCK(&ctx->volfile_lock);
+ gf_msg(THIS->name, GF_LOG_ERROR, 0, glusterfsd_msg_39,
+ "Unable to create temporary file: %s", template);
+ ret = -1;
+ goto out;
+ }
+
+ /* Calling unlink so that when the file is closed or program
+ * terminates the temporary file is deleted.
+ */
+ ret = sys_unlink(template);
+ if (ret < 0) {
+ gf_msg(THIS->name, GF_LOG_INFO, 0, glusterfsd_msg_39,
+ "Unable to delete temporary file: %s", template);
+ ret = 0;
+ }
+
+ tmpfp = fdopen(tmp_fd, "w+b");
+ if (!tmpfp) {
+ ret = -1;
+ goto unlock;
+ }
+
+ fwrite(volfile, size, 1, tmpfp);
+ fflush(tmpfp);
+ if (ferror(tmpfp)) {
+ ret = -1;
+ goto unlock;
+ }
+
+ if (!volfile_tmp) {
+ /* There is no checksum in the list, which means simple attach
+ * the volfile
+ */
+ ret = glusterfs_process_svc_attach_volfp(ctx, tmpfp, volfile_id,
+ sha256_hash);
+ goto unlock;
+ }
+ ret = glusterfs_mux_volfile_reconfigure(tmpfp, ctx, volfile_obj,
+ sha256_hash);
+ if (ret < 0) {
+ gf_msg_debug("glusterfsd-mgmt", EINVAL, "Reconfigure failed !!");
+ }
+ }
+unlock:
+ UNLOCK(&ctx->volfile_lock);
+out:
+ if (tmpfp)
+ fclose(tmpfp);
+ else if (tmp_fd != -1)
+ sys_close(tmp_fd);
+ return ret;
+}
+
+int
mgmt_cbk_event(struct rpc_clnt *rpc, void *mydata, void *data)
{
return 0;
@@ -966,6 +1069,110 @@ glusterfs_handle_attach(rpcsvc_request_t *req)
}
int
+glusterfs_handle_svc_attach(rpcsvc_request_t *req)
+{
+ int32_t ret = -1;
+ gd1_mgmt_brick_op_req xlator_req = {
+ 0,
+ };
+ xlator_t *this = NULL;
+ glusterfs_ctx_t *ctx = NULL;
+
+ GF_ASSERT(req);
+ this = THIS;
+ GF_ASSERT(this);
+
+ ctx = this->ctx;
+ ret = xdr_to_generic(req->msg[0], &xlator_req,
+ (xdrproc_t)xdr_gd1_mgmt_brick_op_req);
+
+ if (ret < 0) {
+ /*failed to decode msg;*/
+ req->rpc_err = GARBAGE_ARGS;
+ goto out;
+ }
+ gf_msg(THIS->name, GF_LOG_INFO, 0, glusterfsd_msg_41,
+ "received attach "
+ "request for volfile-id=%s",
+ xlator_req.name);
+ ret = 0;
+
+ if (ctx->active) {
+ ret = mgmt_process_volfile(xlator_req.input.input_val,
+ xlator_req.input.input_len, xlator_req.name);
+ } else {
+ gf_msg(this->name, GF_LOG_WARNING, EINVAL, glusterfsd_msg_42,
+ "got attach for %s but no active graph", xlator_req.name);
+ }
+out:
+ if (xlator_req.input.input_val)
+ free(xlator_req.input.input_val);
+ if (xlator_req.name)
+ free(xlator_req.name);
+ glusterfs_translator_info_response_send(req, ret, NULL, NULL);
+ return 0;
+}
+
+int
+glusterfs_handle_svc_detach(rpcsvc_request_t *req)
+{
+ gd1_mgmt_brick_op_req xlator_req = {
+ 0,
+ };
+ ssize_t ret;
+ glusterfs_ctx_t *ctx = NULL;
+ gf_volfile_t *volfile_obj = NULL;
+ gf_volfile_t *volfile_tmp = NULL;
+
+ ret = xdr_to_generic(req->msg[0], &xlator_req,
+ (xdrproc_t)xdr_gd1_mgmt_brick_op_req);
+ if (ret < 0) {
+ req->rpc_err = GARBAGE_ARGS;
+ return -1;
+ }
+ ctx = glusterfsd_ctx;
+
+ LOCK(&ctx->volfile_lock);
+ {
+ list_for_each_entry(volfile_obj, &ctx->volfile_list, volfile_list)
+ {
+ if (!strcmp(xlator_req.name, volfile_obj->vol_id)) {
+ volfile_tmp = volfile_obj;
+ break;
+ }
+ }
+
+ if (!volfile_tmp) {
+ UNLOCK(&ctx->volfile_lock);
+ gf_msg(THIS->name, GF_LOG_ERROR, 0, glusterfsd_msg_41,
+ "can't detach %s - not found", xlator_req.name);
+ /*
+ * Used to be -ENOENT. However, the caller asked us to
+ * make sure it's down and if it's already down that's
+ * good enough.
+ */
+ ret = 0;
+ goto out;
+ }
+ ret = glusterfs_process_svc_detach(ctx, volfile_tmp);
+ if (ret) {
+ UNLOCK(&ctx->volfile_lock);
+ gf_msg("glusterfsd-mgmt", GF_LOG_ERROR, EINVAL, glusterfsd_msg_41,
+ "Could not detach "
+ "old graph. Aborting the reconfiguration operation");
+ goto out;
+ }
+ }
+ UNLOCK(&ctx->volfile_lock);
+out:
+ glusterfs_terminate_response_send(req, ret);
+ free(xlator_req.name);
+ xlator_req.name = NULL;
+
+ return 0;
+}
+
+int
glusterfs_handle_dump_metrics(rpcsvc_request_t *req)
{
int32_t ret = -1;
@@ -1849,6 +2056,13 @@ rpcsvc_actor_t glusterfs_actors[GLUSTERD_BRICK_MAXVALUE] = {
[GLUSTERD_DUMP_METRICS] = {"DUMP METRICS", GLUSTERD_DUMP_METRICS,
glusterfs_handle_dump_metrics, NULL, 0, DRC_NA},
+
+ [GLUSTERD_SVC_ATTACH] = {"ATTACH CLIENT", GLUSTERD_SVC_ATTACH,
+ glusterfs_handle_svc_attach, NULL, 0, DRC_NA},
+
+ [GLUSTERD_SVC_DETACH] = {"DETACH CLIENT", GLUSTERD_SVC_DETACH,
+ glusterfs_handle_svc_detach, NULL, 0, DRC_NA},
+
};
struct rpcsvc_program glusterfs_mop_prog = {
@@ -1996,14 +2210,17 @@ mgmt_getspec_cbk(struct rpc_req *req, struct iovec *iov, int count,
}
volfile:
- ret = 0;
size = rsp.op_ret;
+ volfile_id = frame->local;
+ if (mgmt_is_multiplexed_daemon(ctx->cmd_args.process_name)) {
+ ret = mgmt_process_volfile((const char *)rsp.spec, size, volfile_id);
+ goto post_graph_mgmt;
+ }
+ ret = 0;
glusterfs_compute_sha256((const unsigned char *)rsp.spec, size,
sha256_hash);
- volfile_id = frame->local;
-
LOCK(&ctx->volfile_lock);
{
locked = 1;
@@ -2105,6 +2322,7 @@ volfile:
}
INIT_LIST_HEAD(&volfile_tmp->volfile_list);
+ volfile_tmp->graph = ctx->active;
list_add(&volfile_tmp->volfile_list, &ctx->volfile_list);
snprintf(volfile_tmp->vol_id, sizeof(volfile_tmp->vol_id), "%s",
volfile_id);
@@ -2116,6 +2334,7 @@ volfile:
locked = 0;
+post_graph_mgmt:
if (!is_mgmt_rpc_reconnect) {
need_emancipate = 1;
glusterfs_mgmt_pmap_signin(ctx);
@@ -2269,10 +2488,21 @@ glusterfs_volfile_fetch(glusterfs_ctx_t *ctx)
{
xlator_t *server_xl = NULL;
xlator_list_t *trav;
- int ret;
+ gf_volfile_t *volfile_obj = NULL;
+ int ret = 0;
LOCK(&ctx->volfile_lock);
{
+ if (ctx->active &&
+ mgmt_is_multiplexed_daemon(ctx->cmd_args.process_name)) {
+ list_for_each_entry(volfile_obj, &ctx->volfile_list, volfile_list)
+ {
+ ret |= glusterfs_volfile_fetch_one(ctx, volfile_obj->vol_id);
+ }
+ UNLOCK(&ctx->volfile_lock);
+ return ret;
+ }
+
if (ctx->active) {
server_xl = ctx->active->first;
if (strcmp(server_xl->type, "protocol/server") != 0) {
diff --git a/glusterfsd/src/glusterfsd.c b/glusterfsd/src/glusterfsd.c
index c983882..3aa89ca 100644
--- a/glusterfsd/src/glusterfsd.c
+++ b/glusterfsd/src/glusterfsd.c
@@ -2589,24 +2589,6 @@ out:
#endif
int
-glusterfs_graph_fini(glusterfs_graph_t *graph)
-{
- xlator_t *trav = NULL;
-
- trav = graph->first;
-
- while (trav) {
- if (trav->init_succeeded) {
- trav->fini(trav);
- trav->init_succeeded = 0;
- }
- trav = trav->next;
- }
-
- return 0;
-}
-
-int
glusterfs_process_volfp(glusterfs_ctx_t *ctx, FILE *fp)
{
glusterfs_graph_t *graph = NULL;
diff --git a/libglusterfs/src/defaults-tmpl.c b/libglusterfs/src/defaults-tmpl.c
index 5bf64e8..82e7f78 100644
--- a/libglusterfs/src/defaults-tmpl.c
+++ b/libglusterfs/src/defaults-tmpl.c
@@ -127,6 +127,12 @@ default_notify(xlator_t *this, int32_t event, void *data, ...)
GF_UNUSED int ret = 0;
xlator_t *victim = data;
+ glusterfs_graph_t *graph = NULL;
+
+ GF_VALIDATE_OR_GOTO("notify", this, out);
+ graph = this->graph;
+ GF_VALIDATE_OR_GOTO(this->name, graph, out);
+
switch (event) {
case GF_EVENT_PARENT_UP:
case GF_EVENT_PARENT_DOWN: {
@@ -159,6 +165,17 @@ default_notify(xlator_t *this, int32_t event, void *data, ...)
xlator_notify(parent->xlator, event, this, NULL);
parent = parent->next;
}
+
+ if (event == GF_EVENT_CHILD_DOWN &&
+ !(this->ctx && this->ctx->master) && (graph->top == this)) {
+ /* Make sure this is not a daemon with master xlator */
+ pthread_mutex_lock(&graph->mutex);
+ {
+ graph->used = 0;
+ pthread_cond_broadcast(&graph->child_down_cond);
+ }
+ pthread_mutex_unlock(&graph->mutex);
+ }
} break;
case GF_EVENT_UPCALL: {
xlator_list_t *parent = this->parents;
@@ -205,7 +222,7 @@ default_notify(xlator_t *this, int32_t event, void *data, ...)
* nothing to do with readability.
*/
}
-
+out:
return 0;
}
diff --git a/libglusterfs/src/glusterfs/glusterfs.h b/libglusterfs/src/glusterfs/glusterfs.h
index 7c6af09..deec5ba 100644
--- a/libglusterfs/src/glusterfs/glusterfs.h
+++ b/libglusterfs/src/glusterfs/glusterfs.h
@@ -590,6 +590,10 @@ struct _glusterfs_graph {
int used; /* Should be set when fuse gets
first CHILD_UP */
uint32_t volfile_checksum;
+ void *last_xl; /* Stores the last xl of the graph, as of now only populated
+ in client multiplexed code path */
+ pthread_mutex_t mutex;
+ pthread_cond_t child_down_cond; /* for broadcasting CHILD_DOWN */
};
typedef struct _glusterfs_graph glusterfs_graph_t;
@@ -732,6 +736,7 @@ typedef struct {
char volfile_checksum[SHA256_DIGEST_LENGTH];
char vol_id[NAME_MAX + 1];
struct list_head volfile_list;
+ glusterfs_graph_t *graph;
} gf_volfile_t;
@@ -815,4 +820,6 @@ gf_free_mig_locks(lock_migration_info_t *locks);
int
glusterfs_read_secure_access_file(void);
+int
+glusterfs_graph_fini(glusterfs_graph_t *graph);
#endif /* _GLUSTERFS_H */
diff --git a/libglusterfs/src/glusterfs/libglusterfs-messages.h b/libglusterfs/src/glusterfs/libglusterfs-messages.h
index 1b72f6d..ea2aa60 100644
--- a/libglusterfs/src/glusterfs/libglusterfs-messages.h
+++ b/libglusterfs/src/glusterfs/libglusterfs-messages.h
@@ -109,6 +109,8 @@ GLFS_MSGID(
LG_MSG_PTHREAD_ATTR_INIT_FAILED, LG_MSG_INVALID_INODE_LIST,
LG_MSG_COMPACT_FAILED, LG_MSG_COMPACT_STATUS, LG_MSG_UTIMENSAT_FAILED,
LG_MSG_PTHREAD_NAMING_FAILED, LG_MSG_SYSCALL_RETURNS_WRONG,
- LG_MSG_XXH64_TO_GFID_FAILED);
+ LG_MSG_XXH64_TO_GFID_FAILED, LG_MSG_ASYNC_WARNING, LG_MSG_ASYNC_FAILURE,
+ LG_MSG_GRAPH_CLEANUP_FAILED, LG_MSG_GRAPH_SETUP_FAILED,
+ LG_MSG_GRAPH_DETACH_STARTED, LG_MSG_GRAPH_ATTACH_FAILED);
#endif /* !_LG_MESSAGES_H_ */
diff --git a/libglusterfs/src/glusterfs/xlator.h b/libglusterfs/src/glusterfs/xlator.h
index 7002657..06152ec 100644
--- a/libglusterfs/src/glusterfs/xlator.h
+++ b/libglusterfs/src/glusterfs/xlator.h
@@ -1087,4 +1087,7 @@ handle_default_options(xlator_t *xl, dict_t *options);
void
gluster_graph_take_reference(xlator_t *tree);
+
+gf_boolean_t
+mgmt_is_multiplexed_daemon(char *name);
#endif /* _XLATOR_H */
diff --git a/libglusterfs/src/graph.c b/libglusterfs/src/graph.c
index bb5e67a..a492dd8 100644
--- a/libglusterfs/src/graph.c
+++ b/libglusterfs/src/graph.c
@@ -114,6 +114,53 @@ out:
return cert_depth;
}
+xlator_t *
+glusterfs_get_last_xlator(glusterfs_graph_t *graph)
+{
+ xlator_t *trav = graph->first;
+ if (!trav)
+ return NULL;
+
+ while (trav->next)
+ trav = trav->next;
+
+ return trav;
+}
+
+xlator_t *
+glusterfs_mux_xlator_unlink(xlator_t *pxl, xlator_t *cxl)
+{
+ xlator_list_t *unlink = NULL;
+ xlator_list_t *prev = NULL;
+ xlator_list_t **tmp = NULL;
+ xlator_t *next_child = NULL;
+ xlator_t *xl = NULL;
+
+ for (tmp = &pxl->children; *tmp; tmp = &(*tmp)->next) {
+ if ((*tmp)->xlator == cxl) {
+ unlink = *tmp;
+ *tmp = (*tmp)->next;
+ if (*tmp)
+ next_child = (*tmp)->xlator;
+ break;
+ }
+ prev = *tmp;
+ }
+
+ if (!prev)
+ xl = pxl;
+ else if (prev->xlator)
+ xl = prev->xlator->graph->last_xl;
+
+ if (xl)
+ xl->next = next_child;
+ if (next_child)
+ next_child->prev = xl;
+
+ GF_FREE(unlink);
+ return next_child;
+}
+
int
glusterfs_xlator_link(xlator_t *pxl, xlator_t *cxl)
{
@@ -1092,6 +1139,8 @@ glusterfs_graph_destroy_residual(glusterfs_graph_t *graph)
ret = xlator_tree_free_memacct(graph->first);
list_del_init(&graph->list);
+ pthread_mutex_destroy(&graph->mutex);
+ pthread_cond_destroy(&graph->child_down_cond);
GF_FREE(graph);
return ret;
@@ -1134,6 +1183,25 @@ out:
}
int
+glusterfs_graph_fini(glusterfs_graph_t *graph)
+{
+ xlator_t *trav = NULL;
+
+ trav = graph->first;
+
+ while (trav) {
+ if (trav->init_succeeded) {
+ trav->cleanup_starting = 1;
+ trav->fini(trav);
+ trav->init_succeeded = 0;
+ }
+ trav = trav->next;
+ }
+
+ return 0;
+}
+
+int
glusterfs_graph_attach(glusterfs_graph_t *orig_graph, char *path,
glusterfs_graph_t **newgraph)
{
@@ -1256,3 +1324,386 @@ glusterfs_graph_attach(glusterfs_graph_t *orig_graph, char *path,
return 0;
}
+int
+glusterfs_muxsvc_cleanup_parent(glusterfs_ctx_t *ctx,
+ glusterfs_graph_t *parent_graph)
+{
+ if (parent_graph) {
+ if (parent_graph->first) {
+ xlator_destroy(parent_graph->first);
+ }
+ ctx->active = NULL;
+ GF_FREE(parent_graph);
+ parent_graph = NULL;
+ }
+ return 0;
+}
+
+void *
+glusterfs_graph_cleanup(void *arg)
+{
+ glusterfs_graph_t *graph = NULL;
+ glusterfs_ctx_t *ctx = THIS->ctx;
+ int ret = -1;
+ graph = arg;
+
+ if (!graph)
+ return NULL;
+
+ /* To destroy the graph, fitst sent a GF_EVENT_PARENT_DOWN
+ * Then wait for GF_EVENT_CHILD_DOWN to get on the top
+ * xl. Once we have GF_EVENT_CHILD_DOWN event, then proceed
+ * to fini.
+ *
+ * During fini call, this will take a last unref on rpc and
+ * rpc_transport_object.
+ */
+ if (graph->first)
+ default_notify(graph->first, GF_EVENT_PARENT_DOWN, graph->first);
+
+ ret = pthread_mutex_lock(&graph->mutex);
+ if (ret != 0) {
+ gf_msg("glusterfs", GF_LOG_ERROR, EAGAIN, LG_MSG_GRAPH_CLEANUP_FAILED,
+ "Failed to aquire a lock");
+ goto out;
+ }
+ /* check and wait for CHILD_DOWN for top xlator*/
+ while (graph->used) {
+ ret = pthread_cond_wait(&graph->child_down_cond, &graph->mutex);
+ if (ret != 0)
+ gf_msg("glusterfs", GF_LOG_INFO, 0, LG_MSG_GRAPH_CLEANUP_FAILED,
+ "cond wait failed ");
+ }
+
+ ret = pthread_mutex_unlock(&graph->mutex);
+ if (ret != 0) {
+ gf_msg("glusterfs", GF_LOG_ERROR, EAGAIN, LG_MSG_GRAPH_CLEANUP_FAILED,
+ "Failed to release a lock");
+ }
+
+ /* Though we got a child down on top xlator, we have to wait until
+ * all the notifier to exit. Because there should not be any threads
+ * that access xl variables.
+ */
+ pthread_mutex_lock(&ctx->notify_lock);
+ {
+ while (ctx->notifying)
+ pthread_cond_wait(&ctx->notify_cond, &ctx->notify_lock);
+ }
+ pthread_mutex_unlock(&ctx->notify_lock);
+
+ glusterfs_graph_fini(graph);
+ glusterfs_graph_destroy(graph);
+out:
+ return NULL;
+}
+
+glusterfs_graph_t *
+glusterfs_muxsvc_setup_parent_graph(glusterfs_ctx_t *ctx, char *name,
+ char *type)
+{
+ glusterfs_graph_t *parent_graph = NULL;
+ xlator_t *ixl = NULL;
+ int ret = -1;
+ parent_graph = GF_CALLOC(1, sizeof(*parent_graph),
+ gf_common_mt_glusterfs_graph_t);
+ if (!parent_graph)
+ goto out;
+
+ INIT_LIST_HEAD(&parent_graph->list);
+
+ ctx->active = parent_graph;
+ ixl = GF_CALLOC(1, sizeof(*ixl), gf_common_mt_xlator_t);
+ if (!ixl)
+ goto out;
+
+ ixl->ctx = ctx;
+ ixl->graph = parent_graph;
+ ixl->options = dict_new();
+ if (!ixl->options)
+ goto out;
+
+ ixl->name = gf_strdup(name);
+ if (!ixl->name)
+ goto out;
+
+ ixl->is_autoloaded = 1;
+
+ if (xlator_set_type(ixl, type) == -1) {
+ gf_msg("glusterfs", GF_LOG_ERROR, EINVAL, LG_MSG_GRAPH_SETUP_FAILED,
+ "%s (%s) set type failed", name, type);
+ goto out;
+ }
+
+ glusterfs_graph_set_first(parent_graph, ixl);
+ parent_graph->top = ixl;
+ ixl = NULL;
+
+ gettimeofday(&parent_graph->dob, NULL);
+ fill_uuid(parent_graph->graph_uuid, 128);
+ parent_graph->id = ctx->graph_id++;
+ ret = 0;
+out:
+ if (ixl)
+ xlator_destroy(ixl);
+
+ if (ret) {
+ glusterfs_muxsvc_cleanup_parent(ctx, parent_graph);
+ parent_graph = NULL;
+ }
+ return parent_graph;
+}
+
+int
+glusterfs_process_svc_detach(glusterfs_ctx_t *ctx, gf_volfile_t *volfile_obj)
+{
+ xlator_t *last_xl = NULL;
+ glusterfs_graph_t *graph = NULL;
+ glusterfs_graph_t *parent_graph = NULL;
+ pthread_t clean_graph = {
+ 0,
+ };
+ int ret = -1;
+ xlator_t *xl = NULL;
+
+ if (!ctx || !ctx->active || !volfile_obj)
+ goto out;
+ parent_graph = ctx->active;
+ graph = volfile_obj->graph;
+ if (graph && graph->first)
+ xl = graph->first;
+
+ last_xl = graph->last_xl;
+ if (last_xl)
+ last_xl->next = NULL;
+ if (!xl || xl->cleanup_starting)
+ goto out;
+
+ xl->cleanup_starting = 1;
+ gf_msg("mgmt", GF_LOG_INFO, 0, LG_MSG_GRAPH_DETACH_STARTED,
+ "detaching child %s", volfile_obj->vol_id);
+
+ list_del_init(&volfile_obj->volfile_list);
+ glusterfs_mux_xlator_unlink(parent_graph->top, xl);
+ parent_graph->last_xl = glusterfs_get_last_xlator(parent_graph);
+ parent_graph->xl_count -= graph->xl_count;
+ parent_graph->leaf_count -= graph->leaf_count;
+ default_notify(xl, GF_EVENT_PARENT_DOWN, xl);
+ parent_graph->id++;
+ ret = 0;
+out:
+ if (!ret) {
+ list_del_init(&volfile_obj->volfile_list);
+ if (graph) {
+ ret = gf_thread_create_detached(
+ &clean_graph, glusterfs_graph_cleanup, graph, "graph_clean");
+ if (ret) {
+ gf_msg("glusterfs", GF_LOG_ERROR, EINVAL,
+ LG_MSG_GRAPH_CLEANUP_FAILED,
+ "%s failed to create clean "
+ "up thread",
+ volfile_obj->vol_id);
+ ret = 0;
+ }
+ }
+ GF_FREE(volfile_obj);
+ }
+ return ret;
+}
+
+int
+glusterfs_process_svc_attach_volfp(glusterfs_ctx_t *ctx, FILE *fp,
+ char *volfile_id, char *checksum)
+{
+ glusterfs_graph_t *graph = NULL;
+ glusterfs_graph_t *parent_graph = NULL;
+ glusterfs_graph_t *clean_graph = NULL;
+ int ret = -1;
+ xlator_t *xl = NULL;
+ xlator_t *last_xl = NULL;
+ gf_volfile_t *volfile_obj = NULL;
+ pthread_t thread_id = {
+ 0,
+ };
+
+ if (!ctx)
+ goto out;
+ parent_graph = ctx->active;
+ graph = glusterfs_graph_construct(fp);
+ if (!graph) {
+ gf_msg("glusterfsd", GF_LOG_ERROR, EINVAL, LG_MSG_GRAPH_ATTACH_FAILED,
+ "failed to construct the graph");
+ goto out;
+ }
+ graph->last_xl = glusterfs_get_last_xlator(graph);
+
+ for (xl = graph->first; xl; xl = xl->next) {
+ if (strcmp(xl->type, "mount/fuse") == 0) {
+ gf_msg("glusterfsd", GF_LOG_ERROR, EINVAL,
+ LG_MSG_GRAPH_ATTACH_FAILED,
+ "fuse xlator cannot be specified in volume file");
+ goto out;
+ }
+ }
+
+ graph->leaf_count = glusterfs_count_leaves(glusterfs_root(graph));
+ xl = graph->first;
+ /* TODO memory leaks everywhere need to free graph in case of error */
+ if (glusterfs_graph_prepare(graph, ctx, xl->name)) {
+ gf_msg("glusterfsd", GF_LOG_WARNING, EINVAL, LG_MSG_GRAPH_ATTACH_FAILED,
+ "failed to prepare graph for xlator %s", xl->name);
+ ret = -1;
+ goto out;
+ } else if (glusterfs_graph_init(graph)) {
+ gf_msg("glusterfsd", GF_LOG_WARNING, EINVAL, LG_MSG_GRAPH_ATTACH_FAILED,
+ "failed to initialize graph for xlator %s", xl->name);
+ ret = -1;
+ goto out;
+ } else if (glusterfs_graph_parent_up(graph)) {
+ gf_msg("glusterfsd", GF_LOG_WARNING, EINVAL, LG_MSG_GRAPH_ATTACH_FAILED,
+ "failed to link the graphs for xlator %s ", xl->name);
+ ret = -1;
+ goto out;
+ }
+
+ if (!parent_graph) {
+ parent_graph = glusterfs_muxsvc_setup_parent_graph(ctx, "glustershd",
+ "debug/io-stats");
+ if (!parent_graph)
+ goto out;
+ ((xlator_t *)parent_graph->top)->next = xl;
+ clean_graph = parent_graph;
+ } else {
+ last_xl = parent_graph->last_xl;
+ if (last_xl)
+ last_xl->next = xl;
+ xl->prev = last_xl;
+ }
+ parent_graph->last_xl = graph->last_xl;
+
+ ret = glusterfs_xlator_link(parent_graph->top, xl);
+ if (ret) {
+ gf_msg("graph", GF_LOG_ERROR, 0, LG_MSG_EVENT_NOTIFY_FAILED,
+ "parent up notification failed");
+ goto out;
+ }
+ parent_graph->xl_count += graph->xl_count;
+ parent_graph->leaf_count += graph->leaf_count;
+ parent_graph->id++;
+
+ if (!volfile_obj) {
+ volfile_obj = GF_CALLOC(1, sizeof(gf_volfile_t), gf_common_volfile_t);
+ if (!volfile_obj) {
+ ret = -1;
+ goto out;
+ }
+ }
+
+ graph->used = 1;
+ parent_graph->id++;
+ list_add(&graph->list, &ctx->graphs);
+ INIT_LIST_HEAD(&volfile_obj->volfile_list);
+ volfile_obj->graph = graph;
+ snprintf(volfile_obj->vol_id, sizeof(volfile_obj->vol_id), "%s",
+ volfile_id);
+ memcpy(volfile_obj->volfile_checksum, checksum,
+ sizeof(volfile_obj->volfile_checksum));
+ list_add_tail(&volfile_obj->volfile_list, &ctx->volfile_list);
+
+ gf_log_dump_graph(fp, graph);
+ graph = NULL;
+
+ ret = 0;
+out:
+ if (ret) {
+ if (graph) {
+ gluster_graph_take_reference(graph->first);
+ ret = gf_thread_create_detached(&thread_id, glusterfs_graph_cleanup,
+ graph, "graph_clean");
+ if (ret) {
+ gf_msg("glusterfs", GF_LOG_ERROR, EINVAL,
+ LG_MSG_GRAPH_CLEANUP_FAILED,
+ "%s failed to create clean "
+ "up thread",
+ volfile_id);
+ ret = 0;
+ }
+ }
+ if (clean_graph)
+ glusterfs_muxsvc_cleanup_parent(ctx, clean_graph);
+ }
+ return ret;
+}
+
+int
+glusterfs_mux_volfile_reconfigure(FILE *newvolfile_fp, glusterfs_ctx_t *ctx,
+ gf_volfile_t *volfile_obj, char *checksum)
+{
+ glusterfs_graph_t *oldvolfile_graph = NULL;
+ glusterfs_graph_t *newvolfile_graph = NULL;
+
+ int ret = -1;
+
+ if (!ctx) {
+ gf_msg("glusterfsd-mgmt", GF_LOG_ERROR, 0, LG_MSG_CTX_NULL,
+ "ctx is NULL");
+ goto out;
+ }
+
+ /* Change the message id */
+ if (!volfile_obj) {
+ gf_msg("glusterfsd-mgmt", GF_LOG_ERROR, 0, LG_MSG_CTX_NULL,
+ "failed to get volfile object");
+ goto out;
+ }
+
+ oldvolfile_graph = volfile_obj->graph;
+ if (!oldvolfile_graph) {
+ goto out;
+ }
+
+ newvolfile_graph = glusterfs_graph_construct(newvolfile_fp);
+
+ if (!newvolfile_graph) {
+ goto out;
+ }
+ newvolfile_graph->last_xl = glusterfs_get_last_xlator(newvolfile_graph);
+
+ glusterfs_graph_prepare(newvolfile_graph, ctx, newvolfile_graph->first);
+
+ if (!is_graph_topology_equal(oldvolfile_graph, newvolfile_graph)) {
+ ret = glusterfs_process_svc_detach(ctx, volfile_obj);
+ if (ret) {
+ gf_msg("glusterfsd-mgmt", GF_LOG_ERROR, EINVAL,
+ LG_MSG_GRAPH_CLEANUP_FAILED,
+ "Could not detach "
+ "old graph. Aborting the reconfiguration operation");
+ goto out;
+ }
+ ret = glusterfs_process_svc_attach_volfp(ctx, newvolfile_fp,
+ volfile_obj->vol_id, checksum);
+ goto out;
+ }
+
+ gf_msg_debug("glusterfsd-mgmt", 0,
+ "Only options have changed in the"
+ " new graph");
+
+ ret = glusterfs_graph_reconfigure(oldvolfile_graph, newvolfile_graph);
+ if (ret) {
+ gf_msg_debug("glusterfsd-mgmt", 0,
+ "Could not reconfigure "
+ "new options in old graph");
+ goto out;
+ }
+ memcpy(volfile_obj->volfile_checksum, checksum,
+ sizeof(volfile_obj->volfile_checksum));
+
+ ret = 0;
+out:
+
+ if (newvolfile_graph)
+ glusterfs_graph_destroy(newvolfile_graph);
+
+ return ret;
+}
diff --git a/libglusterfs/src/graph.y b/libglusterfs/src/graph.y
index 5b92985..c60ff38 100644
--- a/libglusterfs/src/graph.y
+++ b/libglusterfs/src/graph.y
@@ -542,6 +542,9 @@ glusterfs_graph_new ()
INIT_LIST_HEAD (&graph->list);
+ pthread_mutex_init(&graph->mutex, NULL);
+ pthread_cond_init(&graph->child_down_cond, NULL);
+
gettimeofday (&graph->dob, NULL);
return graph;
diff --git a/libglusterfs/src/libglusterfs.sym b/libglusterfs/src/libglusterfs.sym
index e33d5cf..fa2025e 100644
--- a/libglusterfs/src/libglusterfs.sym
+++ b/libglusterfs/src/libglusterfs.sym
@@ -1154,3 +1154,8 @@ gf_changelog_register_generic
gf_gfid_generate_from_xxh64
find_xlator_option_in_cmd_args_t
gf_d_type_from_ia_type
+glusterfs_graph_fini
+glusterfs_process_svc_attach_volfp
+glusterfs_mux_volfile_reconfigure
+glusterfs_process_svc_detach
+mgmt_is_multiplexed_daemon
diff --git a/libglusterfs/src/statedump.c b/libglusterfs/src/statedump.c
index d18b50f..0cf80c0 100644
--- a/libglusterfs/src/statedump.c
+++ b/libglusterfs/src/statedump.c
@@ -810,7 +810,8 @@ gf_proc_dump_info(int signum, glusterfs_ctx_t *ctx)
if (!ctx)
goto out;
- if (ctx && ctx->active) {
+ if (!mgmt_is_multiplexed_daemon(ctx->cmd_args.process_name) &&
+ (ctx && ctx->active)) {
top = ctx->active->first;
for (trav_p = &top->children; *trav_p; trav_p = &(*trav_p)->next) {
brick_count++;
diff --git a/libglusterfs/src/xlator.c b/libglusterfs/src/xlator.c
index dc1e887..5d6f8d2 100644
--- a/libglusterfs/src/xlator.c
+++ b/libglusterfs/src/xlator.c
@@ -1463,3 +1463,19 @@ gluster_graph_take_reference(xlator_t *tree)
}
return;
}
+
+gf_boolean_t
+mgmt_is_multiplexed_daemon(char *name)
+{
+ const char *mux_daemons[] = {"glustershd", NULL};
+ int i;
+
+ if (!name)
+ return _gf_false;
+
+ for (i = 0; mux_daemons[i]; i++) {
+ if (!strcmp(name, mux_daemons[i]))
+ return _gf_true;
+ }
+ return _gf_false;
+}
diff --git a/rpc/rpc-lib/src/protocol-common.h b/rpc/rpc-lib/src/protocol-common.h
index 779878f..7275d75 100644
--- a/rpc/rpc-lib/src/protocol-common.h
+++ b/rpc/rpc-lib/src/protocol-common.h
@@ -245,6 +245,8 @@ enum glusterd_brick_procnum {
GLUSTERD_NODE_BITROT,
GLUSTERD_BRICK_ATTACH,
GLUSTERD_DUMP_METRICS,
+ GLUSTERD_SVC_ATTACH,
+ GLUSTERD_SVC_DETACH,
GLUSTERD_BRICK_MAXVALUE,
};
diff --git a/tests/basic/glusterd/heald.t b/tests/basic/glusterd/heald.t
index ca112ad..7dae3c3 100644
--- a/tests/basic/glusterd/heald.t
+++ b/tests/basic/glusterd/heald.t
@@ -7,11 +7,16 @@
# Covers enable/disable at the moment. Will be enhanced later to include
# the other commands as well.
+function is_pid_running {
+ local pid=$1
+ num=`ps auxww | grep glustershd | grep $pid | grep -v grep | wc -l`
+ echo $num
+}
+
cleanup;
TEST glusterd
TEST pidof glusterd
-volfile=$(gluster system:: getwd)"/glustershd/glustershd-server.vol"
#Commands should fail when volume doesn't exist
TEST ! $CLI volume heal non-existent-volume enable
TEST ! $CLI volume heal non-existent-volume disable
@@ -20,51 +25,55 @@ TEST ! $CLI volume heal non-existent-volume disable
# volumes
TEST $CLI volume create dist $H0:$B0/dist
TEST $CLI volume start dist
-TEST "[ -z $(get_shd_process_pid)]"
+TEST "[ -z $(get_shd_process_pid dist)]"
TEST ! $CLI volume heal dist enable
TEST ! $CLI volume heal dist disable
# Commands should work on replicate/disperse volume.
TEST $CLI volume create r2 replica 2 $H0:$B0/r2_0 $H0:$B0/r2_1
-TEST "[ -z $(get_shd_process_pid)]"
+TEST "[ -z $(get_shd_process_pid r2)]"
TEST $CLI volume start r2
-EXPECT_WITHIN $PROCESS_UP_TIMEOUT "[0-9][0-9]*" get_shd_process_pid
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "[0-9][0-9]*" get_shd_process_pid r2
TEST $CLI volume heal r2 enable
EXPECT "enable" volume_option r2 "cluster.self-heal-daemon"
-EXPECT "enable" volgen_volume_option $volfile r2-replicate-0 cluster replicate self-heal-daemon
-EXPECT_WITHIN $PROCESS_UP_TIMEOUT "[0-9][0-9]*" get_shd_process_pid
+volfiler2=$(gluster system:: getwd)"/vols/r2/r2-shd.vol"
+EXPECT "enable" volgen_volume_option $volfiler2 r2-replicate-0 cluster replicate self-heal-daemon
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "[0-9][0-9]*" get_shd_process_pid r2
+pid=$( get_shd_process_pid r2 )
TEST $CLI volume heal r2 disable
EXPECT "disable" volume_option r2 "cluster.self-heal-daemon"
-EXPECT "disable" volgen_volume_option $volfile r2-replicate-0 cluster replicate self-heal-daemon
-EXPECT_WITHIN $PROCESS_UP_TIMEOUT "[0-9][0-9]*" get_shd_process_pid
+EXPECT "disable" volgen_volume_option $volfiler2 r2-replicate-0 cluster replicate self-heal-daemon
+EXPECT "1" is_pid_running $pid
# Commands should work on disperse volume.
TEST $CLI volume create ec2 disperse 3 redundancy 1 $H0:$B0/ec2_0 $H0:$B0/ec2_1 $H0:$B0/ec2_2
TEST $CLI volume start ec2
-EXPECT_WITHIN $PROCESS_UP_TIMEOUT "[0-9][0-9]*" get_shd_process_pid
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "[0-9][0-9]*" get_shd_process_pid ec2
TEST $CLI volume heal ec2 enable
EXPECT "enable" volume_option ec2 "cluster.disperse-self-heal-daemon"
-EXPECT "enable" volgen_volume_option $volfile ec2-disperse-0 cluster disperse self-heal-daemon
-EXPECT_WITHIN $PROCESS_UP_TIMEOUT "[0-9][0-9]*" get_shd_process_pid
+volfileec2=$(gluster system:: getwd)"/vols/ec2/ec2-shd.vol"
+EXPECT "enable" volgen_volume_option $volfileec2 ec2-disperse-0 cluster disperse self-heal-daemon
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "[0-9][0-9]*" get_shd_process_pid ec2
+pid=$(get_shd_process_pid ec2)
TEST $CLI volume heal ec2 disable
EXPECT "disable" volume_option ec2 "cluster.disperse-self-heal-daemon"
-EXPECT "disable" volgen_volume_option $volfile ec2-disperse-0 cluster disperse self-heal-daemon
-EXPECT_WITHIN $PROCESS_UP_TIMEOUT "[0-9][0-9]*" get_shd_process_pid
+EXPECT "disable" volgen_volume_option $volfileec2 ec2-disperse-0 cluster disperse self-heal-daemon
+EXPECT "1" is_pid_running $pid
#Check that shd graph is rewritten correctly on volume stop/start
-EXPECT "Y" volgen_volume_exists $volfile ec2-disperse-0 cluster disperse
-EXPECT "Y" volgen_volume_exists $volfile r2-replicate-0 cluster replicate
+EXPECT "Y" volgen_volume_exists $volfileec2 ec2-disperse-0 cluster disperse
+
+EXPECT "Y" volgen_volume_exists $volfiler2 r2-replicate-0 cluster replicate
TEST $CLI volume stop r2
-EXPECT "Y" volgen_volume_exists $volfile ec2-disperse-0 cluster disperse
-EXPECT "N" volgen_volume_exists $volfile r2-replicate-0 cluster replicate
+EXPECT "Y" volgen_volume_exists $volfileec2 ec2-disperse-0 cluster disperse
TEST $CLI volume stop ec2
# When both the volumes are stopped glustershd volfile is not modified just the
# process is stopped
-TEST "[ -z $(get_shd_process_pid) ]"
+TEST "[ -z $(get_shd_process_pid dist) ]"
+TEST "[ -z $(get_shd_process_pid ec2) ]"
TEST $CLI volume start r2
-EXPECT "N" volgen_volume_exists $volfile ec2-disperse-0 cluster disperse
-EXPECT "Y" volgen_volume_exists $volfile r2-replicate-0 cluster replicate
+EXPECT "Y" volgen_volume_exists $volfiler2 r2-replicate-0 cluster replicate
TEST $CLI volume set r2 self-heal-daemon on
TEST $CLI volume set r2 cluster.self-heal-daemon off
diff --git a/tests/bugs/glusterd/reset-brick-and-daemons-follow-quorum.t b/tests/bugs/glusterd/reset-brick-and-daemons-follow-quorum.t
index cdb1a33..e6e65c4 100644
--- a/tests/bugs/glusterd/reset-brick-and-daemons-follow-quorum.t
+++ b/tests/bugs/glusterd/reset-brick-and-daemons-follow-quorum.t
@@ -55,9 +55,9 @@ TEST kill_glusterd 1
#Bring back 1st glusterd
TEST $glusterd_1
-# We need to wait till PROCESS_UP_TIMEOUT and then check shd service does not
-# come up on node 2
-sleep $PROCESS_UP_TIMEOUT
-EXPECT "N" shd_up_status_2
+# We need to wait till PROCESS_UP_TIMEOUT and then check shd service started
+#on node 2, because once glusterd regains quorum, it will restart all volume
+#level daemons
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "Y" shd_up_status_2
cleanup;
diff --git a/tests/volume.rc b/tests/volume.rc
index 87ca958..289b197 100644
--- a/tests/volume.rc
+++ b/tests/volume.rc
@@ -237,11 +237,13 @@ function ec_child_up_count_shd {
}
function get_shd_process_pid {
- ps auxww | grep glusterfs | grep -E "glustershd/glustershd.pid" | awk '{print $2}' | head -1
+ local vol=$1
+ ps auxww | grep "process-name\ glustershd" | awk '{print $2}' | head -1
}
function generate_shd_statedump {
- generate_statedump $(get_shd_process_pid)
+ local vol=$1
+ generate_statedump $(get_shd_process_pid $vol)
}
function generate_nfs_statedump {
diff --git a/xlators/mgmt/glusterd/src/Makefile.am b/xlators/mgmt/glusterd/src/Makefile.am
index 5fe5156..11ae189 100644
--- a/xlators/mgmt/glusterd/src/Makefile.am
+++ b/xlators/mgmt/glusterd/src/Makefile.am
@@ -18,11 +18,12 @@ glusterd_la_SOURCES = glusterd.c glusterd-handler.c glusterd-sm.c \
glusterd-locks.c glusterd-snapshot.c glusterd-mgmt-handler.c \
glusterd-mgmt.c glusterd-peer-utils.c glusterd-statedump.c \
glusterd-snapshot-utils.c glusterd-conn-mgmt.c \
- glusterd-proc-mgmt.c glusterd-svc-mgmt.c glusterd-shd-svc.c \
+ glusterd-proc-mgmt.c glusterd-svc-mgmt.c \
glusterd-nfs-svc.c glusterd-quotad-svc.c glusterd-svc-helper.c \
glusterd-conn-helper.c glusterd-snapd-svc.c glusterd-snapd-svc-helper.c \
glusterd-bitd-svc.c glusterd-scrub-svc.c glusterd-server-quorum.c \
glusterd-reset-brick.c glusterd-tierd-svc.c glusterd-tierd-svc-helper.c \
+ glusterd-shd-svc.c glusterd-shd-svc-helper.c \
glusterd-gfproxyd-svc.c glusterd-gfproxyd-svc-helper.c
@@ -38,11 +39,12 @@ noinst_HEADERS = glusterd.h glusterd-utils.h glusterd-op-sm.h \
glusterd-mgmt.h glusterd-messages.h glusterd-peer-utils.h \
glusterd-statedump.h glusterd-snapshot-utils.h glusterd-geo-rep.h \
glusterd-conn-mgmt.h glusterd-conn-helper.h glusterd-proc-mgmt.h \
- glusterd-svc-mgmt.h glusterd-shd-svc.h glusterd-nfs-svc.h \
+ glusterd-svc-mgmt.h glusterd-nfs-svc.h \
glusterd-quotad-svc.h glusterd-svc-helper.h glusterd-snapd-svc.h \
glusterd-snapd-svc-helper.h glusterd-rcu.h glusterd-bitd-svc.h \
glusterd-scrub-svc.h glusterd-server-quorum.h glusterd-errno.h \
glusterd-tierd-svc.h glusterd-tierd-svc-helper.h \
+ glusterd-shd-svc.h glusterd-shd-svc-helper.h \
glusterd-gfproxyd-svc.h glusterd-gfproxyd-svc-helper.h \
$(CONTRIBDIR)/userspace-rcu/rculist-extra.h
diff --git a/xlators/mgmt/glusterd/src/glusterd-brick-ops.c b/xlators/mgmt/glusterd/src/glusterd-brick-ops.c
index ad9a572..042a805 100644
--- a/xlators/mgmt/glusterd/src/glusterd-brick-ops.c
+++ b/xlators/mgmt/glusterd/src/glusterd-brick-ops.c
@@ -2863,7 +2863,7 @@ glusterd_op_remove_brick(dict_t *dict, char **op_errstr)
}
if (start_remove && volinfo->status == GLUSTERD_STATUS_STARTED) {
- ret = glusterd_svcs_reconfigure();
+ ret = glusterd_svcs_reconfigure(volinfo);
if (ret) {
gf_msg(this->name, GF_LOG_WARNING, 0, GD_MSG_NFS_RECONF_FAIL,
"Unable to reconfigure NFS-Server");
diff --git a/xlators/mgmt/glusterd/src/glusterd-conn-mgmt.c b/xlators/mgmt/glusterd/src/glusterd-conn-mgmt.c
index e80e152..052438c 100644
--- a/xlators/mgmt/glusterd/src/glusterd-conn-mgmt.c
+++ b/xlators/mgmt/glusterd/src/glusterd-conn-mgmt.c
@@ -132,3 +132,45 @@ glusterd_conn_build_socket_filepath(char *rundir, uuid_t uuid, char *socketpath,
glusterd_set_socket_filepath(sockfilepath, socketpath, len);
return 0;
}
+
+int
+__glusterd_muxsvc_conn_common_notify(struct rpc_clnt *rpc, void *mydata,
+ rpc_clnt_event_t event, void *data)
+{
+ glusterd_conf_t *conf = THIS->private;
+ glusterd_svc_proc_t *mux_proc = mydata;
+ int ret = -1;
+
+ /* Silently ignoring this error, exactly like the current
+ * implementation */
+ if (!mux_proc)
+ return 0;
+
+ if (event == RPC_CLNT_DESTROY) {
+ /*RPC_CLNT_DESTROY will only called after mux_proc detached from the
+ * list. So it is safe to call without lock. Processing
+ * RPC_CLNT_DESTROY under a lock will lead to deadlock.
+ */
+ if (mux_proc->data) {
+ glusterd_volinfo_unref(mux_proc->data);
+ mux_proc->data = NULL;
+ }
+ GF_FREE(mux_proc);
+ ret = 0;
+ } else {
+ pthread_mutex_lock(&conf->attach_lock);
+ {
+ ret = mux_proc->notify(mux_proc, event);
+ }
+ pthread_mutex_unlock(&conf->attach_lock);
+ }
+ return ret;
+}
+
+int
+glusterd_muxsvc_conn_common_notify(struct rpc_clnt *rpc, void *mydata,
+ rpc_clnt_event_t event, void *data)
+{
+ return glusterd_big_locked_notify(rpc, mydata, event, data,
+ __glusterd_muxsvc_conn_common_notify);
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-conn-mgmt.h b/xlators/mgmt/glusterd/src/glusterd-conn-mgmt.h
index 602c0ba..d1c4607 100644
--- a/xlators/mgmt/glusterd/src/glusterd-conn-mgmt.h
+++ b/xlators/mgmt/glusterd/src/glusterd-conn-mgmt.h
@@ -43,9 +43,11 @@ glusterd_conn_disconnect(glusterd_conn_t *conn);
int
glusterd_conn_common_notify(struct rpc_clnt *rpc, void *mydata,
rpc_clnt_event_t event, void *data);
+int
+glusterd_muxsvc_conn_common_notify(struct rpc_clnt *rpc, void *mydata,
+ rpc_clnt_event_t event, void *data);
int32_t
glusterd_conn_build_socket_filepath(char *rundir, uuid_t uuid, char *socketpath,
int len);
-
#endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-gfproxyd-svc.c b/xlators/mgmt/glusterd/src/glusterd-gfproxyd-svc.c
index f9c8617..b01fd4d 100644
--- a/xlators/mgmt/glusterd/src/glusterd-gfproxyd-svc.c
+++ b/xlators/mgmt/glusterd/src/glusterd-gfproxyd-svc.c
@@ -370,6 +370,7 @@ int
glusterd_gfproxydsvc_restart()
{
glusterd_volinfo_t *volinfo = NULL;
+ glusterd_volinfo_t *tmp = NULL;
int ret = -1;
xlator_t *this = THIS;
glusterd_conf_t *conf = NULL;
@@ -380,7 +381,7 @@ glusterd_gfproxydsvc_restart()
conf = this->private;
GF_VALIDATE_OR_GOTO(this->name, conf, out);
- cds_list_for_each_entry(volinfo, &conf->volumes, vol_list)
+ cds_list_for_each_entry_safe(volinfo, tmp, &conf->volumes, vol_list)
{
/* Start per volume gfproxyd svc */
if (volinfo->status == GLUSTERD_STATUS_STARTED) {
diff --git a/xlators/mgmt/glusterd/src/glusterd-handler.c b/xlators/mgmt/glusterd/src/glusterd-handler.c
index 528993c..1cb9013 100644
--- a/xlators/mgmt/glusterd/src/glusterd-handler.c
+++ b/xlators/mgmt/glusterd/src/glusterd-handler.c
@@ -5928,6 +5928,11 @@ glusterd_get_state(rpcsvc_request_t *req, dict_t *dict)
GF_FREE(rebal_data);
+ fprintf(fp, "Volume%d.shd_svc.online_status: %s\n", count,
+ volinfo->shd.svc.online ? "Online" : "Offline");
+ fprintf(fp, "Volume%d.shd_svc.inited: %s\n", count,
+ volinfo->shd.svc.inited ? "True" : "False");
+
if (volinfo->type == GF_CLUSTER_TYPE_TIER) {
ret = glusterd_volume_get_hot_tier_type_str(volinfo,
&hot_tier_type_str);
@@ -5997,12 +6002,6 @@ glusterd_get_state(rpcsvc_request_t *req, dict_t *dict)
fprintf(fp, "\n[Services]\n");
- if (priv->shd_svc.inited) {
- fprintf(fp, "svc%d.name: %s\n", ++count, priv->shd_svc.name);
- fprintf(fp, "svc%d.online_status: %s\n\n", count,
- priv->shd_svc.online ? "Online" : "Offline");
- }
-
if (priv->nfs_svc.inited) {
fprintf(fp, "svc%d.name: %s\n", ++count, priv->nfs_svc.name);
fprintf(fp, "svc%d.online_status: %s\n\n", count,
diff --git a/xlators/mgmt/glusterd/src/glusterd-handshake.c b/xlators/mgmt/glusterd/src/glusterd-handshake.c
index 5599a63..1ba58c3 100644
--- a/xlators/mgmt/glusterd/src/glusterd-handshake.c
+++ b/xlators/mgmt/glusterd/src/glusterd-handshake.c
@@ -30,6 +30,7 @@
#include "rpcsvc.h"
#include "rpc-common-xdr.h"
#include "glusterd-gfproxyd-svc-helper.h"
+#include "glusterd-shd-svc-helper.h"
extern struct rpc_clnt_program gd_peer_prog;
extern struct rpc_clnt_program gd_mgmt_prog;
@@ -328,6 +329,26 @@ build_volfile_path(char *volume_id, char *path, size_t path_len,
goto out;
}
+ volid_ptr = strstr(volume_id, "shd/");
+ if (volid_ptr) {
+ volid_ptr = strchr(volid_ptr, '/');
+ if (!volid_ptr) {
+ ret = -1;
+ goto out;
+ }
+ volid_ptr++;
+
+ ret = glusterd_volinfo_find(volid_ptr, &volinfo);
+ if (ret == -1) {
+ gf_log(this->name, GF_LOG_ERROR, "Couldn't find volinfo");
+ goto out;
+ }
+
+ glusterd_svc_build_shd_volfile_path(volinfo, path, path_len);
+ ret = 0;
+ goto out;
+ }
+
volid_ptr = strstr(volume_id, "/snaps/");
if (volid_ptr) {
ret = get_snap_volname_and_volinfo(volid_ptr, &volname, &volinfo);
diff --git a/xlators/mgmt/glusterd/src/glusterd-mem-types.h b/xlators/mgmt/glusterd/src/glusterd-mem-types.h
index 7a784db..17052ce 100644
--- a/xlators/mgmt/glusterd/src/glusterd-mem-types.h
+++ b/xlators/mgmt/glusterd/src/glusterd-mem-types.h
@@ -51,6 +51,7 @@ typedef enum gf_gld_mem_types_ {
gf_gld_mt_missed_snapinfo_t,
gf_gld_mt_snap_create_args_t,
gf_gld_mt_glusterd_brick_proc_t,
+ gf_gld_mt_glusterd_svc_proc_t,
gf_gld_mt_end,
} gf_gld_mem_types_t;
#endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-messages.h b/xlators/mgmt/glusterd/src/glusterd-messages.h
index c7b3ca8..424e15f 100644
--- a/xlators/mgmt/glusterd/src/glusterd-messages.h
+++ b/xlators/mgmt/glusterd/src/glusterd-messages.h
@@ -298,6 +298,8 @@ GLFS_MSGID(
GD_MSG_LOCALTIME_LOGGING_ENABLE, GD_MSG_LOCALTIME_LOGGING_DISABLE,
GD_MSG_PORTS_EXHAUSTED, GD_MSG_CHANGELOG_GET_FAIL,
GD_MSG_MANAGER_FUNCTION_FAILED, GD_MSG_NFS_GANESHA_DISABLED,
- GD_MSG_GANESHA_NOT_RUNNING, GD_MSG_DAEMON_LOG_LEVEL_VOL_OPT_VALIDATE_FAIL);
+ GD_MSG_GANESHA_NOT_RUNNING, GD_MSG_DAEMON_LOG_LEVEL_VOL_OPT_VALIDATE_FAIL,
+ GD_MSG_SHD_START_FAIL, GD_MSG_SHD_OBJ_GET_FAIL, GD_MSG_SVC_ATTACH_FAIL,
+ GD_MSG_ATTACH_INFO, GD_MSG_DETACH_INFO, GD_MSG_SVC_DETACH_FAIL);
#endif /* !_GLUSTERD_MESSAGES_H_ */
diff --git a/xlators/mgmt/glusterd/src/glusterd-op-sm.c b/xlators/mgmt/glusterd/src/glusterd-op-sm.c
index df8a6ab..95f9707 100644
--- a/xlators/mgmt/glusterd/src/glusterd-op-sm.c
+++ b/xlators/mgmt/glusterd/src/glusterd-op-sm.c
@@ -44,6 +44,7 @@
#include "glusterd-snapshot-utils.h"
#include "glusterd-svc-mgmt.h"
#include "glusterd-svc-helper.h"
+#include "glusterd-shd-svc-helper.h"
#include "glusterd-shd-svc.h"
#include "glusterd-nfs-svc.h"
#include "glusterd-quotad-svc.h"
@@ -2223,6 +2224,11 @@ glusterd_options_reset(glusterd_volinfo_t *volinfo, char *key,
if (ret)
goto out;
+ svc = &(volinfo->shd.svc);
+ ret = svc->reconfigure(volinfo);
+ if (ret)
+ goto out;
+
ret = glusterd_create_volfiles_and_notify_services(volinfo);
if (ret) {
gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_VOLFILE_CREATE_FAIL,
@@ -2237,7 +2243,7 @@ glusterd_options_reset(glusterd_volinfo_t *volinfo, char *key,
goto out;
if (GLUSTERD_STATUS_STARTED == volinfo->status) {
- ret = glusterd_svcs_reconfigure();
+ ret = glusterd_svcs_reconfigure(volinfo);
if (ret)
goto out;
}
@@ -2693,6 +2699,11 @@ glusterd_op_set_all_volume_options(xlator_t *this, dict_t *dict,
if (ret)
goto out;
+ svc = &(volinfo->shd.svc);
+ ret = svc->reconfigure(volinfo);
+ if (ret)
+ goto out;
+
ret = glusterd_create_volfiles_and_notify_services(volinfo);
if (ret) {
gf_msg(this->name, GF_LOG_ERROR, 0,
@@ -2706,7 +2717,7 @@ glusterd_op_set_all_volume_options(xlator_t *this, dict_t *dict,
}
}
if (svcs_reconfigure) {
- ret = glusterd_svcs_reconfigure();
+ ret = glusterd_svcs_reconfigure(NULL);
if (ret) {
gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_SVC_RESTART_FAIL,
"Unable to restart "
@@ -3091,6 +3102,11 @@ glusterd_op_set_volume(dict_t *dict, char **errstr)
if (ret)
goto out;
+ svc = &(volinfo->shd.svc);
+ ret = svc->reconfigure(volinfo);
+ if (ret)
+ goto out;
+
ret = glusterd_create_volfiles_and_notify_services(volinfo);
if (ret) {
gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_VOLFILE_CREATE_FAIL,
@@ -3106,7 +3122,7 @@ glusterd_op_set_volume(dict_t *dict, char **errstr)
goto out;
if (GLUSTERD_STATUS_STARTED == volinfo->status) {
- ret = glusterd_svcs_reconfigure();
+ ret = glusterd_svcs_reconfigure(volinfo);
if (ret) {
gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_SVC_RESTART_FAIL,
"Unable to restart services");
@@ -3139,6 +3155,11 @@ glusterd_op_set_volume(dict_t *dict, char **errstr)
if (ret)
goto out;
+ svc = &(volinfo->shd.svc);
+ ret = svc->reconfigure(volinfo);
+ if (ret)
+ goto out;
+
ret = glusterd_create_volfiles_and_notify_services(volinfo);
if (ret) {
gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_VOLFILE_CREATE_FAIL,
@@ -3154,7 +3175,7 @@ glusterd_op_set_volume(dict_t *dict, char **errstr)
goto out;
if (GLUSTERD_STATUS_STARTED == volinfo->status) {
- ret = glusterd_svcs_reconfigure();
+ ret = glusterd_svcs_reconfigure(volinfo);
if (ret) {
gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_SVC_RESTART_FAIL,
"Unable to restart services");
@@ -3361,7 +3382,7 @@ glusterd_op_stats_volume(dict_t *dict, char **op_errstr, dict_t *rsp_dict)
goto out;
if (GLUSTERD_STATUS_STARTED == volinfo->status) {
- ret = glusterd_svcs_reconfigure();
+ ret = glusterd_svcs_reconfigure(volinfo);
if (ret)
goto out;
}
@@ -3644,14 +3665,6 @@ glusterd_op_status_volume(dict_t *dict, char **op_errstr, dict_t *rsp_dict)
other_count++;
node_count++;
- } else if ((cmd & GF_CLI_STATUS_SHD) != 0) {
- ret = glusterd_add_node_to_dict(priv->shd_svc.name, rsp_dict, 0,
- vol_opts);
- if (ret)
- goto out;
- other_count++;
- node_count++;
-
} else if ((cmd & GF_CLI_STATUS_QUOTAD) != 0) {
ret = glusterd_add_node_to_dict(priv->quotad_svc.name, rsp_dict, 0,
vol_opts);
@@ -3685,6 +3698,12 @@ glusterd_op_status_volume(dict_t *dict, char **op_errstr, dict_t *rsp_dict)
goto out;
other_count++;
node_count++;
+ } else if ((cmd & GF_CLI_STATUS_SHD) != 0) {
+ ret = glusterd_add_shd_to_dict(volinfo, rsp_dict, other_index);
+ if (ret)
+ goto out;
+ other_count++;
+ node_count++;
} else if ((cmd & GF_CLI_STATUS_BRICK) != 0) {
ret = dict_get_strn(dict, "brick", SLEN("brick"), &brick);
if (ret)
@@ -3747,6 +3766,19 @@ glusterd_op_status_volume(dict_t *dict, char **op_errstr, dict_t *rsp_dict)
node_count++;
}
+ if (glusterd_is_shd_compatible_volume(volinfo)) {
+ shd_enabled = gd_is_self_heal_enabled(volinfo, vol_opts);
+ if (shd_enabled) {
+ ret = glusterd_add_shd_to_dict(volinfo, rsp_dict,
+ other_index);
+ if (ret)
+ goto out;
+ other_count++;
+ other_index++;
+ node_count++;
+ }
+ }
+
nfs_disabled = dict_get_str_boolean(vol_opts, NFS_DISABLE_MAP_KEY,
_gf_false);
if (!nfs_disabled) {
@@ -3759,18 +3791,6 @@ glusterd_op_status_volume(dict_t *dict, char **op_errstr, dict_t *rsp_dict)
node_count++;
}
- if (glusterd_is_shd_compatible_volume(volinfo))
- shd_enabled = gd_is_self_heal_enabled(volinfo, vol_opts);
- if (shd_enabled) {
- ret = glusterd_add_node_to_dict(priv->shd_svc.name, rsp_dict,
- other_index, vol_opts);
- if (ret)
- goto out;
- other_count++;
- node_count++;
- other_index++;
- }
-
if (glusterd_is_volume_quota_enabled(volinfo)) {
ret = glusterd_add_node_to_dict(priv->quotad_svc.name, rsp_dict,
other_index, vol_opts);
@@ -6875,16 +6895,18 @@ glusterd_shd_select_brick_xlator(dict_t *dict, gf_xl_afr_op_t heal_op,
int ret = -1;
glusterd_conf_t *priv = NULL;
xlator_t *this = NULL;
+ glusterd_svc_t *svc = NULL;
this = THIS;
GF_ASSERT(this);
priv = this->private;
GF_ASSERT(priv);
+ svc = &(volinfo->shd.svc);
switch (heal_op) {
case GF_SHD_OP_INDEX_SUMMARY:
case GF_SHD_OP_STATISTICS_HEAL_COUNT:
- if (!priv->shd_svc.online) {
+ if (!svc->online) {
if (!rsp_dict) {
gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_OPCTX_NULL,
"Received "
@@ -6905,7 +6927,7 @@ glusterd_shd_select_brick_xlator(dict_t *dict, gf_xl_afr_op_t heal_op,
break;
case GF_SHD_OP_STATISTICS_HEAL_COUNT_PER_REPLICA:
- if (!priv->shd_svc.online) {
+ if (!svc->online) {
if (!rsp_dict) {
gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_OPCTX_NULL,
"Received "
@@ -7040,7 +7062,7 @@ glusterd_bricks_select_heal_volume(dict_t *dict, char **op_errstr,
ret = -1;
goto out;
} else {
- pending_node->node = &(priv->shd_svc);
+ pending_node->node = &(volinfo->shd.svc);
pending_node->type = GD_NODE_SHD;
cds_list_add_tail(&pending_node->list, selected);
pending_node = NULL;
@@ -7174,6 +7196,7 @@ glusterd_bricks_select_status_volume(dict_t *dict, char **op_errstr,
glusterd_pending_node_t *pending_node = NULL;
xlator_t *this = NULL;
glusterd_conf_t *priv = NULL;
+ glusterd_svc_t *svc = NULL;
GF_ASSERT(dict);
@@ -7269,7 +7292,8 @@ glusterd_bricks_select_status_volume(dict_t *dict, char **op_errstr,
ret = 0;
} else if ((cmd & GF_CLI_STATUS_SHD) != 0) {
- if (!priv->shd_svc.online) {
+ svc = &(volinfo->shd.svc);
+ if (!svc->online) {
ret = -1;
gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_SELF_HEALD_DISABLED,
"Self-heal daemon is not running");
@@ -7281,7 +7305,7 @@ glusterd_bricks_select_status_volume(dict_t *dict, char **op_errstr,
ret = -1;
goto out;
}
- pending_node->node = &(priv->shd_svc);
+ pending_node->node = svc;
pending_node->type = GD_NODE_SHD;
pending_node->index = 0;
cds_list_add_tail(&pending_node->list, selected);
diff --git a/xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.c b/xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.c
new file mode 100644
index 0000000..9196758
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.c
@@ -0,0 +1,140 @@
+/*
+ Copyright (c) 2016 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+*/
+
+#include "glusterd.h"
+#include "glusterd-utils.h"
+#include "glusterd-shd-svc-helper.h"
+#include "glusterd-messages.h"
+#include "glusterd-volgen.h"
+
+void
+glusterd_svc_build_shd_socket_filepath(glusterd_volinfo_t *volinfo, char *path,
+ int path_len)
+{
+ char sockfilepath[PATH_MAX] = {
+ 0,
+ };
+ char rundir[PATH_MAX] = {
+ 0,
+ };
+ int32_t len = 0;
+ glusterd_conf_t *priv = THIS->private;
+
+ if (!priv)
+ return;
+
+ GLUSTERD_GET_SHD_RUNDIR(rundir, volinfo, priv);
+ len = snprintf(sockfilepath, sizeof(sockfilepath), "%s/run-%s", rundir,
+ uuid_utoa(MY_UUID));
+ if ((len < 0) || (len >= sizeof(sockfilepath))) {
+ sockfilepath[0] = 0;
+ }
+
+ glusterd_set_socket_filepath(sockfilepath, path, path_len);
+}
+
+void
+glusterd_svc_build_shd_pidfile(glusterd_volinfo_t *volinfo, char *path,
+ int path_len)
+{
+ char rundir[PATH_MAX] = {
+ 0,
+ };
+ glusterd_conf_t *priv = THIS->private;
+
+ if (!priv)
+ return;
+
+ GLUSTERD_GET_SHD_RUNDIR(rundir, volinfo, priv);
+
+ snprintf(path, path_len, "%s/%s-shd.pid", rundir, volinfo->volname);
+}
+
+void
+glusterd_svc_build_shd_volfile_path(glusterd_volinfo_t *volinfo, char *path,
+ int path_len)
+{
+ char workdir[PATH_MAX] = {
+ 0,
+ };
+ glusterd_conf_t *priv = THIS->private;
+
+ if (!priv)
+ return;
+
+ GLUSTERD_GET_VOLUME_DIR(workdir, volinfo, priv);
+
+ snprintf(path, path_len, "%s/%s-shd.vol", workdir, volinfo->volname);
+}
+
+void
+glusterd_svc_build_shd_logdir(char *logdir, char *volname, size_t len)
+{
+ snprintf(logdir, len, "%s/shd/%s", DEFAULT_LOG_FILE_DIRECTORY, volname);
+}
+
+void
+glusterd_svc_build_shd_logfile(char *logfile, char *logdir, size_t len)
+{
+ snprintf(logfile, len, "%s/shd.log", logdir);
+}
+
+void
+glusterd_shd_svcproc_cleanup(glusterd_shdsvc_t *shd)
+{
+ glusterd_svc_proc_t *svc_proc = NULL;
+ glusterd_svc_t *svc = NULL;
+ glusterd_conf_t *conf = NULL;
+ gf_boolean_t need_unref = _gf_false;
+ rpc_clnt_t *rpc = NULL;
+
+ conf = THIS->private;
+ if (!conf)
+ return;
+
+ GF_VALIDATE_OR_GOTO(THIS->name, conf, out);
+ GF_VALIDATE_OR_GOTO(THIS->name, shd, out);
+
+ svc = &shd->svc;
+ shd->attached = _gf_false;
+
+ if (svc->conn.rpc) {
+ rpc_clnt_unref(svc->conn.rpc);
+ svc->conn.rpc = NULL;
+ }
+
+ pthread_mutex_lock(&conf->attach_lock);
+ {
+ svc_proc = svc->svc_proc;
+ svc->svc_proc = NULL;
+ svc->inited = _gf_false;
+ cds_list_del_init(&svc->mux_svc);
+ glusterd_unlink_file(svc->proc.pidfile);
+
+ if (svc_proc && cds_list_empty(&svc_proc->svcs)) {
+ cds_list_del_init(&svc_proc->svc_proc_list);
+ /* We cannot free svc_proc list from here. Because
+ * if there are pending events on the rpc, it will
+ * try to access the corresponding svc_proc, so unrefing
+ * rpc request and then cleaning up the memory is carried
+ * from the notify function upon RPC_CLNT_DESTROY destroy.
+ */
+ need_unref = _gf_true;
+ rpc = svc_proc->rpc;
+ svc_proc->rpc = NULL;
+ }
+ }
+ pthread_mutex_unlock(&conf->attach_lock);
+ /*rpc unref has to be performed outside the lock*/
+ if (need_unref && rpc)
+ rpc_clnt_unref(rpc);
+out:
+ return;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.h b/xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.h
new file mode 100644
index 0000000..c70702c
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.h
@@ -0,0 +1,45 @@
+/*
+ Copyright (c) 2016 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+*/
+
+#ifndef _GLUSTERD_SHD_SVC_HELPER_H_
+#define _GLUSTERD_SHD_SVC_HELPER_H_
+
+#include "glusterd.h"
+#include "glusterd-svc-mgmt.h"
+
+void
+glusterd_svc_build_shd_socket_filepath(glusterd_volinfo_t *volinfo, char *path,
+ int path_len);
+
+void
+glusterd_svc_build_shd_pidfile(glusterd_volinfo_t *volinfo, char *path,
+ int path_len);
+
+void
+glusterd_svc_build_shd_volfile_path(glusterd_volinfo_t *volinfo, char *path,
+ int path_len);
+
+void
+glusterd_svc_build_shd_logdir(char *logdir, char *volname, size_t len);
+
+void
+glusterd_svc_build_shd_logfile(char *logfile, char *logdir, size_t len);
+
+void
+glusterd_shd_svcproc_cleanup(glusterd_shdsvc_t *shd);
+
+int
+glusterd_recover_shd_attach_failure(glusterd_volinfo_t *volinfo,
+ glusterd_svc_t *svc, int flags);
+
+int
+glusterd_shdsvc_create_volfile(glusterd_volinfo_t *volinfo);
+
+#endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-shd-svc.c b/xlators/mgmt/glusterd/src/glusterd-shd-svc.c
index 69e27cb..937ea30 100644
--- a/xlators/mgmt/glusterd/src/glusterd-shd-svc.c
+++ b/xlators/mgmt/glusterd/src/glusterd-shd-svc.c
@@ -13,9 +13,10 @@
#include "glusterd.h"
#include "glusterd-utils.h"
#include "glusterd-volgen.h"
-#include "glusterd-svc-mgmt.h"
#include "glusterd-shd-svc.h"
+#include "glusterd-shd-svc-helper.h"
#include "glusterd-svc-helper.h"
+#include "glusterd-store.h"
#define GD_SHD_PROCESS_NAME "--process-name"
char *shd_svc_name = "glustershd";
@@ -23,27 +24,145 @@ char *shd_svc_name = "glustershd";
void
glusterd_shdsvc_build(glusterd_svc_t *svc)
{
+ int ret = -1;
+ ret = snprintf(svc->name, sizeof(svc->name), "%s", shd_svc_name);
+ if (ret < 0)
+ return;
+
+ CDS_INIT_LIST_HEAD(&svc->mux_svc);
svc->manager = glusterd_shdsvc_manager;
svc->start = glusterd_shdsvc_start;
- svc->stop = glusterd_svc_stop;
+ svc->stop = glusterd_shdsvc_stop;
+ svc->reconfigure = glusterd_shdsvc_reconfigure;
}
int
-glusterd_shdsvc_init(glusterd_svc_t *svc)
+glusterd_shdsvc_init(void *data, glusterd_conn_t *mux_conn,
+ glusterd_svc_proc_t *mux_svc)
{
- return glusterd_svc_init(svc, shd_svc_name);
+ int ret = -1;
+ char rundir[PATH_MAX] = {
+ 0,
+ };
+ char sockpath[PATH_MAX] = {
+ 0,
+ };
+ char pidfile[PATH_MAX] = {
+ 0,
+ };
+ char volfile[PATH_MAX] = {
+ 0,
+ };
+ char logdir[PATH_MAX] = {
+ 0,
+ };
+ char logfile[PATH_MAX] = {
+ 0,
+ };
+ char volfileid[256] = {0};
+ glusterd_svc_t *svc = NULL;
+ glusterd_volinfo_t *volinfo = NULL;
+ glusterd_conf_t *priv = NULL;
+ glusterd_muxsvc_conn_notify_t notify = NULL;
+ xlator_t *this = NULL;
+ char *volfileserver = NULL;
+ int32_t len = 0;
+
+ this = THIS;
+ GF_VALIDATE_OR_GOTO(THIS->name, this, out);
+
+ priv = this->private;
+ GF_VALIDATE_OR_GOTO(this->name, priv, out);
+
+ volinfo = data;
+ GF_VALIDATE_OR_GOTO(this->name, data, out);
+ GF_VALIDATE_OR_GOTO(this->name, mux_svc, out);
+
+ svc = &(volinfo->shd.svc);
+
+ ret = snprintf(svc->name, sizeof(svc->name), "%s", shd_svc_name);
+ if (ret < 0)
+ goto out;
+
+ notify = glusterd_muxsvc_common_rpc_notify;
+ glusterd_store_perform_node_state_store(volinfo);
+
+ GLUSTERD_GET_SHD_RUNDIR(rundir, volinfo, priv);
+ glusterd_svc_create_rundir(rundir);
+
+ glusterd_svc_build_shd_logdir(logdir, volinfo->volname, sizeof(logdir));
+ glusterd_svc_build_shd_logfile(logfile, logdir, sizeof(logfile));
+
+ /* Initialize the connection mgmt */
+ if (mux_conn && mux_svc->rpc) {
+ /* multiplexed svc */
+ svc->conn.frame_timeout = mux_conn->frame_timeout;
+ /* This will be unrefed from glusterd_shd_svcproc_cleanup*/
+ svc->conn.rpc = rpc_clnt_ref(mux_svc->rpc);
+ ret = snprintf(svc->conn.sockpath, sizeof(svc->conn.sockpath), "%s",
+ mux_conn->sockpath);
+ } else {
+ ret = mkdir_p(logdir, 0755, _gf_true);
+ if ((ret == -1) && (EEXIST != errno)) {
+ gf_msg(this->name, GF_LOG_ERROR, errno, GD_MSG_CREATE_DIR_FAILED,
+ "Unable to create logdir %s", logdir);
+ goto out;
+ }
+
+ glusterd_svc_build_shd_socket_filepath(volinfo, sockpath,
+ sizeof(sockpath));
+ ret = glusterd_muxsvc_conn_init(&(svc->conn), mux_svc, sockpath, 600,
+ notify);
+ if (ret)
+ goto out;
+ /* This will be unrefed when the last svcs is detached from the list */
+ if (!mux_svc->rpc)
+ mux_svc->rpc = rpc_clnt_ref(svc->conn.rpc);
+ }
+
+ /* Initialize the process mgmt */
+ glusterd_svc_build_shd_pidfile(volinfo, pidfile, sizeof(pidfile));
+ glusterd_svc_build_shd_volfile_path(volinfo, volfile, PATH_MAX);
+ len = snprintf(volfileid, sizeof(volfileid), "shd/%s", volinfo->volname);
+ if ((len < 0) || (len >= sizeof(volfileid))) {
+ ret = -1;
+ goto out;
+ }
+
+ if (dict_get_strn(this->options, "transport.socket.bind-address",
+ SLEN("transport.socket.bind-address"),
+ &volfileserver) != 0) {
+ volfileserver = "localhost";
+ }
+ ret = glusterd_proc_init(&(svc->proc), shd_svc_name, pidfile, logdir,
+ logfile, volfile, volfileid, volfileserver);
+ if (ret)
+ goto out;
+
+out:
+ gf_msg_debug(this ? this->name : "glusterd", 0, "Returning %d", ret);
+ return ret;
}
-static int
-glusterd_shdsvc_create_volfile()
+int
+glusterd_shdsvc_create_volfile(glusterd_volinfo_t *volinfo)
{
char filepath[PATH_MAX] = {
0,
};
+
int ret = -1;
- glusterd_conf_t *conf = THIS->private;
dict_t *mod_dict = NULL;
+ glusterd_svc_build_shd_volfile_path(volinfo, filepath, PATH_MAX);
+ if (!glusterd_is_shd_compatible_volume(volinfo)) {
+ /* If volfile exist, delete it. This case happens when we
+ * change from replica/ec to distribute.
+ */
+ (void)glusterd_unlink_file(filepath);
+ ret = 0;
+ goto out;
+ }
mod_dict = dict_new();
if (!mod_dict)
goto out;
@@ -64,9 +183,7 @@ glusterd_shdsvc_create_volfile()
if (ret)
goto out;
- glusterd_svc_build_volfile_path(shd_svc_name, conf->workdir, filepath,
- sizeof(filepath));
- ret = glusterd_create_global_volfile(build_shd_graph, filepath, mod_dict);
+ ret = glusterd_shdsvc_generate_volfile(volinfo, filepath, mod_dict);
if (ret) {
gf_msg(THIS->name, GF_LOG_ERROR, 0, GD_MSG_VOLFILE_CREATE_FAIL,
"Failed to create volfile");
@@ -81,26 +198,89 @@ out:
return ret;
}
+gf_boolean_t
+glusterd_svcs_shd_compatible_volumes_stopped(glusterd_svc_t *svc)
+{
+ glusterd_svc_proc_t *svc_proc = NULL;
+ glusterd_shdsvc_t *shd = NULL;
+ glusterd_svc_t *temp_svc = NULL;
+ glusterd_volinfo_t *volinfo = NULL;
+ gf_boolean_t comp = _gf_false;
+ glusterd_conf_t *conf = THIS->private;
+
+ GF_VALIDATE_OR_GOTO("glusterd", conf, out);
+ GF_VALIDATE_OR_GOTO("glusterd", svc, out);
+ pthread_mutex_lock(&conf->attach_lock);
+ {
+ svc_proc = svc->svc_proc;
+ if (!svc_proc)
+ goto unlock;
+ cds_list_for_each_entry(temp_svc, &svc_proc->svcs, mux_svc)
+ {
+ /* Get volinfo->shd from svc object */
+ shd = cds_list_entry(svc, glusterd_shdsvc_t, svc);
+ if (!shd) {
+ gf_msg("glusterd", GF_LOG_ERROR, 0, GD_MSG_SHD_OBJ_GET_FAIL,
+ "Failed to get shd object "
+ "from shd service");
+ goto unlock;
+ }
+
+ /* Get volinfo from shd */
+ volinfo = cds_list_entry(shd, glusterd_volinfo_t, shd);
+ if (!volinfo) {
+ gf_msg("glusterd", GF_LOG_ERROR, 0, GD_MSG_VOLINFO_GET_FAIL,
+ "Failed to get volinfo from "
+ "from shd");
+ goto unlock;
+ }
+ if (!glusterd_is_shd_compatible_volume(volinfo))
+ continue;
+ if (volinfo->status == GLUSTERD_STATUS_STARTED)
+ goto unlock;
+ }
+ comp = _gf_true;
+ }
+unlock:
+ pthread_mutex_unlock(&conf->attach_lock);
+out:
+ return comp;
+}
+
int
glusterd_shdsvc_manager(glusterd_svc_t *svc, void *data, int flags)
{
- int ret = 0;
+ int ret = -1;
glusterd_volinfo_t *volinfo = NULL;
- if (!svc->inited) {
- ret = glusterd_shdsvc_init(svc);
- if (ret) {
- gf_msg(THIS->name, GF_LOG_ERROR, 0, GD_MSG_FAILED_INIT_SHDSVC,
- "Failed to init shd "
- "service");
- goto out;
- } else {
- svc->inited = _gf_true;
- gf_msg_debug(THIS->name, 0, "shd service initialized");
+ volinfo = data;
+ GF_VALIDATE_OR_GOTO("glusterd", svc, out);
+ GF_VALIDATE_OR_GOTO("glusterd", volinfo, out);
+
+ if (volinfo)
+ glusterd_volinfo_ref(volinfo);
+
+ ret = glusterd_shdsvc_create_volfile(volinfo);
+ if (ret)
+ goto out;
+
+ if (!glusterd_is_shd_compatible_volume(volinfo)) {
+ ret = 0;
+ if (svc->inited) {
+ /* This means glusterd was running for this volume and now
+ * it was converted to a non-shd volume. So just stop the shd
+ */
+ ret = svc->stop(svc, SIGTERM);
}
+ goto out;
}
- volinfo = data;
+ ret = glusterd_shd_svc_mux_init(volinfo, svc);
+ if (ret) {
+ gf_msg(THIS->name, GF_LOG_ERROR, 0, GD_MSG_FAILED_INIT_SHDSVC,
+ "Failed to init shd service");
+ goto out;
+ }
/* If all the volumes are stopped or all shd compatible volumes
* are stopped then stop the service if:
@@ -110,31 +290,26 @@ glusterd_shdsvc_manager(glusterd_svc_t *svc, void *data, int flags)
* - volinfo is NULL or
* - volinfo is present and volume is shd compatible
*/
- if (glusterd_are_all_volumes_stopped() ||
- glusterd_all_shd_compatible_volumes_stopped()) {
- if (!(volinfo && !glusterd_is_shd_compatible_volume(volinfo))) {
- ret = svc->stop(svc, SIGTERM);
- }
- } else {
- if (!(volinfo && !glusterd_is_shd_compatible_volume(volinfo))) {
- ret = glusterd_shdsvc_create_volfile();
- if (ret)
- goto out;
-
- ret = svc->stop(svc, SIGTERM);
- if (ret)
- goto out;
+ if (glusterd_svcs_shd_compatible_volumes_stopped(svc)) {
+ /* TODO
+ * Take a lock and detach all svc's to stop the process
+ * also reset the init flag
+ */
+ ret = svc->stop(svc, SIGTERM);
+ } else if (volinfo) {
+ ret = svc->stop(svc, SIGTERM);
+ if (ret)
+ goto out;
+ if (volinfo->status == GLUSTERD_STATUS_STARTED) {
ret = svc->start(svc, flags);
if (ret)
goto out;
-
- ret = glusterd_conn_connect(&(svc->conn));
- if (ret)
- goto out;
}
}
out:
+ if (volinfo)
+ glusterd_volinfo_unref(volinfo);
if (ret)
gf_event(EVENT_SVC_MANAGER_FAILED, "svc_name=%s", svc->name);
gf_msg_debug(THIS->name, 0, "Returning %d", ret);
@@ -143,7 +318,7 @@ out:
}
int
-glusterd_shdsvc_start(glusterd_svc_t *svc, int flags)
+glusterd_new_shd_svc_start(glusterd_svc_t *svc, int flags)
{
int ret = -1;
char glusterd_uuid_option[PATH_MAX] = {0};
@@ -178,31 +353,136 @@ glusterd_shdsvc_start(glusterd_svc_t *svc, int flags)
goto out;
ret = glusterd_svc_start(svc, flags, cmdline);
+ if (ret)
+ goto out;
+ ret = glusterd_conn_connect(&(svc->conn));
out:
if (cmdline)
dict_unref(cmdline);
+ return ret;
+}
+int
+glusterd_recover_shd_attach_failure(glusterd_volinfo_t *volinfo,
+ glusterd_svc_t *svc, int flags)
+{
+ int ret = -1;
+ glusterd_svc_proc_t *mux_proc = NULL;
+ glusterd_conf_t *conf = NULL;
+
+ conf = THIS->private;
+
+ if (!conf || !volinfo || !svc)
+ return -1;
+ glusterd_shd_svcproc_cleanup(&volinfo->shd);
+ mux_proc = glusterd_svcprocess_new();
+ if (!mux_proc) {
+ return -1;
+ }
+ ret = glusterd_shdsvc_init(volinfo, NULL, mux_proc);
+ if (ret)
+ return -1;
+ pthread_mutex_lock(&conf->attach_lock);
+ {
+ cds_list_add_tail(&mux_proc->svc_proc_list, &conf->shd_procs);
+ svc->svc_proc = mux_proc;
+ cds_list_del_init(&svc->mux_svc);
+ cds_list_add_tail(&svc->mux_svc, &mux_proc->svcs);
+ }
+ pthread_mutex_unlock(&conf->attach_lock);
+
+ ret = glusterd_new_shd_svc_start(svc, flags);
+ if (!ret) {
+ volinfo->shd.attached = _gf_true;
+ }
+ return ret;
+}
+
+int
+glusterd_shdsvc_start(glusterd_svc_t *svc, int flags)
+{
+ int ret = -1;
+ glusterd_shdsvc_t *shd = NULL;
+ glusterd_volinfo_t *volinfo = NULL;
+ glusterd_conf_t *conf = NULL;
+
+ GF_VALIDATE_OR_GOTO("glusterd", svc, out);
+ conf = THIS->private;
+ GF_VALIDATE_OR_GOTO("glusterd", conf, out);
+
+ /* Get volinfo->shd from svc object */
+ shd = cds_list_entry(svc, glusterd_shdsvc_t, svc);
+ if (!shd) {
+ gf_msg("glusterd", GF_LOG_ERROR, 0, GD_MSG_SHD_OBJ_GET_FAIL,
+ "Failed to get shd object "
+ "from shd service");
+ return -1;
+ }
+
+ /* Get volinfo from shd */
+ volinfo = cds_list_entry(shd, glusterd_volinfo_t, shd);
+ if (!volinfo) {
+ gf_msg("glusterd", GF_LOG_ERROR, 0, GD_MSG_VOLINFO_GET_FAIL,
+ "Failed to get volinfo from "
+ "from shd");
+ return -1;
+ }
+
+ if (volinfo->status != GLUSTERD_STATUS_STARTED)
+ return -1;
+
+ glusterd_volinfo_ref(volinfo);
+ if (!svc->inited) {
+ ret = glusterd_shd_svc_mux_init(volinfo, svc);
+ if (ret)
+ goto out;
+ }
+
+ if (shd->attached) {
+ ret = glusterd_attach_svc(svc, volinfo, flags);
+ if (ret) {
+ gf_msg("glusterd", GF_LOG_ERROR, 0, GD_MSG_VOLINFO_GET_FAIL,
+ "Failed to attach shd svc(volume=%s) to pid=%d. Starting"
+ "a new process",
+ volinfo->volname, glusterd_proc_get_pid(&svc->proc));
+ ret = glusterd_recover_shd_attach_failure(volinfo, svc, flags);
+ }
+ goto out;
+ }
+ ret = glusterd_new_shd_svc_start(svc, flags);
+ if (!ret) {
+ shd->attached = _gf_true;
+ }
+out:
+ if (volinfo)
+ glusterd_volinfo_unref(volinfo);
gf_msg_debug(THIS->name, 0, "Returning %d", ret);
return ret;
}
int
-glusterd_shdsvc_reconfigure()
+glusterd_shdsvc_reconfigure(glusterd_volinfo_t *volinfo)
{
int ret = -1;
xlator_t *this = NULL;
- glusterd_conf_t *priv = NULL;
gf_boolean_t identical = _gf_false;
+ dict_t *mod_dict = NULL;
+ glusterd_svc_t *svc = NULL;
this = THIS;
GF_VALIDATE_OR_GOTO("glusterd", this, out);
- priv = this->private;
- GF_VALIDATE_OR_GOTO(this->name, priv, out);
+ if (!volinfo) {
+ /* reconfigure will be called separately*/
+ ret = 0;
+ goto out;
+ }
- if (glusterd_all_shd_compatible_volumes_stopped())
+ glusterd_volinfo_ref(volinfo);
+ svc = &(volinfo->shd.svc);
+ if (glusterd_svcs_shd_compatible_volumes_stopped(svc))
goto manager;
/*
@@ -210,8 +490,42 @@ glusterd_shdsvc_reconfigure()
* and cksum i.e. "character-by-character". If YES, then
* NOTHING has been changed, just return.
*/
- ret = glusterd_svc_check_volfile_identical(priv->shd_svc.name,
- build_shd_graph, &identical);
+
+ if (!glusterd_is_shd_compatible_volume(volinfo)) {
+ if (svc->inited)
+ goto manager;
+
+ /* Nothing to do if not shd compatible */
+ ret = 0;
+ goto out;
+ }
+ mod_dict = dict_new();
+ if (!mod_dict)
+ goto out;
+
+ ret = dict_set_uint32(mod_dict, "cluster.background-self-heal-count", 0);
+ if (ret)
+ goto out;
+
+ ret = dict_set_str(mod_dict, "cluster.data-self-heal", "on");
+ if (ret)
+ goto out;
+
+ ret = dict_set_str(mod_dict, "cluster.metadata-self-heal", "on");
+ if (ret)
+ goto out;
+
+ ret = dict_set_int32(mod_dict, "graph-check", 1);
+ if (ret)
+ goto out;
+
+ ret = dict_set_str(mod_dict, "cluster.entry-self-heal", "on");
+ if (ret)
+ goto out;
+
+ ret = glusterd_volume_svc_check_volfile_identical(
+ "glustershd", mod_dict, volinfo, glusterd_shdsvc_generate_volfile,
+ &identical);
if (ret)
goto out;
@@ -226,8 +540,9 @@ glusterd_shdsvc_reconfigure()
* changed, then inform the xlator to reconfigure the options.
*/
identical = _gf_false; /* RESET the FLAG */
- ret = glusterd_svc_check_topology_identical(priv->shd_svc.name,
- build_shd_graph, &identical);
+ ret = glusterd_volume_svc_check_topology_identical(
+ "glustershd", mod_dict, volinfo, glusterd_shdsvc_generate_volfile,
+ &identical);
if (ret)
goto out;
@@ -235,7 +550,7 @@ glusterd_shdsvc_reconfigure()
* options to shd volfile, so that shd will be reconfigured.
*/
if (identical) {
- ret = glusterd_shdsvc_create_volfile();
+ ret = glusterd_shdsvc_create_volfile(volinfo);
if (ret == 0) { /* Only if above PASSES */
ret = glusterd_fetchspec_notify(THIS);
}
@@ -243,12 +558,129 @@ glusterd_shdsvc_reconfigure()
}
manager:
/*
- * shd volfile's topology has been changed. shd server needs
- * to be RESTARTED to ACT on the changed volfile.
+ * shd volfile's topology has been changed. volfile needs
+ * to be RECONFIGURED to ACT on the changed volfile.
*/
- ret = priv->shd_svc.manager(&(priv->shd_svc), NULL, PROC_START_NO_WAIT);
+ ret = svc->manager(svc, volinfo, PROC_START_NO_WAIT);
out:
+ if (volinfo)
+ glusterd_volinfo_unref(volinfo);
+ if (mod_dict)
+ dict_unref(mod_dict);
gf_msg_debug(this ? this->name : "glusterd", 0, "Returning %d", ret);
return ret;
}
+
+int
+glusterd_shdsvc_restart()
+{
+ glusterd_volinfo_t *volinfo = NULL;
+ glusterd_volinfo_t *tmp = NULL;
+ int ret = -1;
+ xlator_t *this = THIS;
+ glusterd_conf_t *conf = NULL;
+ glusterd_svc_t *svc = NULL;
+
+ GF_VALIDATE_OR_GOTO("glusterd", this, out);
+
+ conf = this->private;
+ GF_VALIDATE_OR_GOTO(this->name, conf, out);
+
+ pthread_mutex_lock(&conf->volume_lock);
+ cds_list_for_each_entry_safe(volinfo, tmp, &conf->volumes, vol_list)
+ {
+ glusterd_volinfo_ref(volinfo);
+ pthread_mutex_unlock(&conf->volume_lock);
+ /* Start per volume shd svc */
+ if (volinfo->status == GLUSTERD_STATUS_STARTED) {
+ svc = &(volinfo->shd.svc);
+ ret = svc->manager(svc, volinfo, PROC_START_NO_WAIT);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_SHD_START_FAIL,
+ "Couldn't start shd for "
+ "vol: %s on restart",
+ volinfo->volname);
+ gf_event(EVENT_SVC_MANAGER_FAILED, "volume=%s;svc_name=%s",
+ volinfo->volname, svc->name);
+ glusterd_volinfo_unref(volinfo);
+ goto out;
+ }
+ }
+ glusterd_volinfo_unref(volinfo);
+ pthread_mutex_lock(&conf->volume_lock);
+ }
+ pthread_mutex_unlock(&conf->volume_lock);
+out:
+ return ret;
+}
+
+int
+glusterd_shdsvc_stop(glusterd_svc_t *svc, int sig)
+{
+ int ret = -1;
+ glusterd_svc_proc_t *svc_proc = NULL;
+ glusterd_shdsvc_t *shd = NULL;
+ glusterd_volinfo_t *volinfo = NULL;
+ gf_boolean_t empty = _gf_false;
+ glusterd_conf_t *conf = NULL;
+ int pid = -1;
+
+ conf = THIS->private;
+ GF_VALIDATE_OR_GOTO("glusterd", svc, out);
+ svc_proc = svc->svc_proc;
+ GF_VALIDATE_OR_GOTO("glusterd", svc_proc, out);
+ GF_VALIDATE_OR_GOTO("glusterd", conf, out);
+
+ /* Get volinfo->shd from svc object */
+ shd = cds_list_entry(svc, glusterd_shdsvc_t, svc);
+ if (!shd) {
+ gf_msg("glusterd", GF_LOG_ERROR, 0, GD_MSG_SHD_OBJ_GET_FAIL,
+ "Failed to get shd object "
+ "from shd service");
+ return -1;
+ }
+
+ /* Get volinfo from shd */
+ volinfo = cds_list_entry(shd, glusterd_volinfo_t, shd);
+ if (!volinfo) {
+ gf_msg("glusterd", GF_LOG_ERROR, 0, GD_MSG_VOLINFO_GET_FAIL,
+ "Failed to get volinfo from "
+ "from shd");
+ return -1;
+ }
+
+ glusterd_volinfo_ref(volinfo);
+ pthread_mutex_lock(&conf->attach_lock);
+ {
+ gf_is_service_running(svc->proc.pidfile, &pid);
+ cds_list_del_init(&svc->mux_svc);
+ empty = cds_list_empty(&svc_proc->svcs);
+ }
+ pthread_mutex_unlock(&conf->attach_lock);
+ if (empty) {
+ /* Unref will happen when destroying the connection */
+ glusterd_volinfo_ref(volinfo);
+ svc_proc->data = volinfo;
+ ret = glusterd_svc_stop(svc, sig);
+ }
+ if (!empty && pid != -1) {
+ ret = glusterd_detach_svc(svc, volinfo, sig);
+ if (ret)
+ gf_msg(THIS->name, GF_LOG_ERROR, 0, GD_MSG_SVC_STOP_FAIL,
+ "shd service is failed to detach volume %s from pid %d",
+ volinfo->volname, glusterd_proc_get_pid(&svc->proc));
+ else
+ gf_msg(THIS->name, GF_LOG_INFO, 0, GD_MSG_SVC_STOP_SUCCESS,
+ "Shd service is detached for volume %s from pid %d",
+ volinfo->volname, glusterd_proc_get_pid(&svc->proc));
+ }
+ svc->online = _gf_false;
+ (void)glusterd_unlink_file((char *)svc->proc.pidfile);
+ glusterd_shd_svcproc_cleanup(shd);
+ ret = 0;
+ glusterd_volinfo_unref(volinfo);
+out:
+ gf_msg_debug(THIS->name, 0, "Returning %d", ret);
+ return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-shd-svc.h b/xlators/mgmt/glusterd/src/glusterd-shd-svc.h
index 775a9d4..55b409f 100644
--- a/xlators/mgmt/glusterd/src/glusterd-shd-svc.h
+++ b/xlators/mgmt/glusterd/src/glusterd-shd-svc.h
@@ -12,12 +12,20 @@
#define _GLUSTERD_SHD_SVC_H_
#include "glusterd-svc-mgmt.h"
+#include "glusterd.h"
+
+typedef struct glusterd_shdsvc_ glusterd_shdsvc_t;
+struct glusterd_shdsvc_ {
+ glusterd_svc_t svc;
+ gf_boolean_t attached;
+};
void
glusterd_shdsvc_build(glusterd_svc_t *svc);
int
-glusterd_shdsvc_init(glusterd_svc_t *svc);
+glusterd_shdsvc_init(void *data, glusterd_conn_t *mux_conn,
+ glusterd_svc_proc_t *svc_proc);
int
glusterd_shdsvc_manager(glusterd_svc_t *svc, void *data, int flags);
@@ -27,4 +35,11 @@ glusterd_shdsvc_start(glusterd_svc_t *svc, int flags);
int
glusterd_shdsvc_reconfigure();
+
+int
+glusterd_shdsvc_restart();
+
+int
+glusterd_shdsvc_stop(glusterd_svc_t *svc, int sig);
+
#endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-sm.c b/xlators/mgmt/glusterd/src/glusterd-sm.c
index 54a7bd1..943b1c6 100644
--- a/xlators/mgmt/glusterd/src/glusterd-sm.c
+++ b/xlators/mgmt/glusterd/src/glusterd-sm.c
@@ -748,6 +748,16 @@ glusterd_peer_detach_cleanup(glusterd_conf_t *priv)
}
}
+ if (glusterd_is_shd_compatible_volume(volinfo)) {
+ svc = &(volinfo->shd.svc);
+ ret = svc->stop(svc, SIGTERM);
+ if (ret) {
+ gf_msg(THIS->name, GF_LOG_ERROR, 0, GD_MSG_SVC_STOP_FAIL,
+ "Failed "
+ "to stop shd daemon service");
+ }
+ }
+
if (glusterd_is_gfproxyd_enabled(volinfo)) {
svc = &(volinfo->gfproxyd.svc);
ret = svc->stop(svc, SIGTERM);
@@ -775,7 +785,7 @@ glusterd_peer_detach_cleanup(glusterd_conf_t *priv)
}
/*Reconfigure all daemon services upon peer detach*/
- ret = glusterd_svcs_reconfigure();
+ ret = glusterd_svcs_reconfigure(NULL);
if (ret) {
gf_msg(THIS->name, GF_LOG_ERROR, 0, GD_MSG_SVC_STOP_FAIL,
"Failed to reconfigure all daemon services.");
diff --git a/xlators/mgmt/glusterd/src/glusterd-snapd-svc.c b/xlators/mgmt/glusterd/src/glusterd-snapd-svc.c
index 56bab07..1da4076 100644
--- a/xlators/mgmt/glusterd/src/glusterd-snapd-svc.c
+++ b/xlators/mgmt/glusterd/src/glusterd-snapd-svc.c
@@ -366,6 +366,7 @@ int
glusterd_snapdsvc_restart()
{
glusterd_volinfo_t *volinfo = NULL;
+ glusterd_volinfo_t *tmp = NULL;
int ret = 0;
xlator_t *this = THIS;
glusterd_conf_t *conf = NULL;
@@ -376,7 +377,7 @@ glusterd_snapdsvc_restart()
conf = this->private;
GF_ASSERT(conf);
- cds_list_for_each_entry(volinfo, &conf->volumes, vol_list)
+ cds_list_for_each_entry_safe(volinfo, tmp, &conf->volumes, vol_list)
{
/* Start per volume snapd svc */
if (volinfo->status == GLUSTERD_STATUS_STARTED) {
diff --git a/xlators/mgmt/glusterd/src/glusterd-statedump.c b/xlators/mgmt/glusterd/src/glusterd-statedump.c
index f5ecde7..69d4cf4 100644
--- a/xlators/mgmt/glusterd/src/glusterd-statedump.c
+++ b/xlators/mgmt/glusterd/src/glusterd-statedump.c
@@ -202,9 +202,6 @@ glusterd_dump_priv(xlator_t *this)
gf_proc_dump_build_key(key, "glusterd", "ping-timeout");
gf_proc_dump_write(key, "%d", priv->ping_timeout);
- gf_proc_dump_build_key(key, "glusterd", "shd.online");
- gf_proc_dump_write(key, "%d", priv->shd_svc.online);
-
gf_proc_dump_build_key(key, "glusterd", "nfs.online");
gf_proc_dump_write(key, "%d", priv->nfs_svc.online);
diff --git a/xlators/mgmt/glusterd/src/glusterd-svc-helper.c b/xlators/mgmt/glusterd/src/glusterd-svc-helper.c
index ca19a75..e42703c 100644
--- a/xlators/mgmt/glusterd/src/glusterd-svc-helper.c
+++ b/xlators/mgmt/glusterd/src/glusterd-svc-helper.c
@@ -7,6 +7,7 @@
later), or the GNU General Public License, version 2 (GPLv2), in all
cases as published by the Free Software Foundation.
*/
+#include <signal.h>
#include <glusterfs/globals.h>
#include <glusterfs/run.h>
@@ -20,12 +21,14 @@
#include "glusterd-bitd-svc.h"
#include "glusterd-tierd-svc.h"
#include "glusterd-tierd-svc-helper.h"
+#include "glusterd-shd-svc-helper.h"
#include "glusterd-scrub-svc.h"
#include "glusterd-svc-helper.h"
#include <glusterfs/syscall.h>
+#include "glusterd-snapshot-utils.h"
int
-glusterd_svcs_reconfigure()
+glusterd_svcs_reconfigure(glusterd_volinfo_t *volinfo)
{
int ret = 0;
xlator_t *this = THIS;
@@ -43,9 +46,11 @@ glusterd_svcs_reconfigure()
goto out;
svc_name = "self-heald";
- ret = glusterd_shdsvc_reconfigure();
- if (ret)
- goto out;
+ if (volinfo) {
+ ret = glusterd_shdsvc_reconfigure(volinfo);
+ if (ret)
+ goto out;
+ }
if (conf->op_version == GD_OP_VERSION_MIN)
goto out;
@@ -69,7 +74,7 @@ out:
}
int
-glusterd_svcs_stop()
+glusterd_svcs_stop(glusterd_volinfo_t *volinfo)
{
int ret = 0;
xlator_t *this = NULL;
@@ -85,14 +90,16 @@ glusterd_svcs_stop()
if (ret)
goto out;
- ret = glusterd_svc_stop(&(priv->shd_svc), SIGTERM);
- if (ret)
- goto out;
-
ret = glusterd_svc_stop(&(priv->quotad_svc), SIGTERM);
if (ret)
goto out;
+ if (volinfo) {
+ ret = glusterd_svc_stop(&(volinfo->shd.svc), PROC_START_NO_WAIT);
+ if (ret)
+ goto out;
+ }
+
ret = glusterd_svc_stop(&(priv->bitd_svc), SIGTERM);
if (ret)
goto out;
@@ -121,12 +128,6 @@ glusterd_svcs_manager(glusterd_volinfo_t *volinfo)
if (ret)
goto out;
- ret = conf->shd_svc.manager(&(conf->shd_svc), volinfo, PROC_START_NO_WAIT);
- if (ret == -EINVAL)
- ret = 0;
- if (ret)
- goto out;
-
if (conf->op_version == GD_OP_VERSION_MIN)
goto out;
@@ -143,6 +144,15 @@ glusterd_svcs_manager(glusterd_volinfo_t *volinfo)
if (ret)
goto out;
+ if (volinfo) {
+ ret = volinfo->shd.svc.manager(&(volinfo->shd.svc), volinfo,
+ PROC_START_NO_WAIT);
+ if (ret == -EINVAL)
+ ret = 0;
+ if (ret)
+ goto out;
+ }
+
ret = conf->scrub_svc.manager(&(conf->scrub_svc), NULL, PROC_START_NO_WAIT);
if (ret == -EINVAL)
ret = 0;
@@ -269,3 +279,678 @@ out:
GF_FREE(tmpvol);
return ret;
}
+
+int
+glusterd_volume_svc_check_volfile_identical(
+ char *svc_name, dict_t *mode_dict, glusterd_volinfo_t *volinfo,
+ glusterd_vol_graph_builder_t builder, gf_boolean_t *identical)
+{
+ char orgvol[PATH_MAX] = {
+ 0,
+ };
+ char *tmpvol = NULL;
+ xlator_t *this = NULL;
+ int ret = -1;
+ int need_unlink = 0;
+ int tmp_fd = -1;
+
+ this = THIS;
+
+ GF_VALIDATE_OR_GOTO(this->name, this, out);
+ GF_VALIDATE_OR_GOTO(this->name, identical, out);
+
+ /* This builds volfile for volume level dameons */
+ glusterd_volume_svc_build_volfile_path(svc_name, volinfo, orgvol,
+ sizeof(orgvol));
+
+ ret = gf_asprintf(&tmpvol, "/tmp/g%s-XXXXXX", svc_name);
+ if (ret < 0) {
+ goto out;
+ }
+
+ /* coverity[secure_temp] mkstemp uses 0600 as the mode and is safe */
+ tmp_fd = mkstemp(tmpvol);
+ if (tmp_fd < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, errno, GD_MSG_FILE_OP_FAILED,
+ "Unable to create temp file"
+ " %s:(%s)",
+ tmpvol, strerror(errno));
+ ret = -1;
+ goto out;
+ }
+
+ need_unlink = 1;
+
+ ret = builder(volinfo, tmpvol, mode_dict);
+ if (ret)
+ goto out;
+
+ ret = glusterd_check_files_identical(orgvol, tmpvol, identical);
+out:
+ if (need_unlink)
+ sys_unlink(tmpvol);
+
+ if (tmpvol != NULL)
+ GF_FREE(tmpvol);
+
+ if (tmp_fd >= 0)
+ sys_close(tmp_fd);
+
+ return ret;
+}
+
+int
+glusterd_volume_svc_check_topology_identical(
+ char *svc_name, dict_t *mode_dict, glusterd_volinfo_t *volinfo,
+ glusterd_vol_graph_builder_t builder, gf_boolean_t *identical)
+{
+ char orgvol[PATH_MAX] = {
+ 0,
+ };
+ char *tmpvol = NULL;
+ glusterd_conf_t *conf = NULL;
+ xlator_t *this = THIS;
+ int ret = -1;
+ int tmpclean = 0;
+ int tmpfd = -1;
+
+ if ((!identical) || (!this) || (!this->private))
+ goto out;
+
+ conf = this->private;
+ GF_VALIDATE_OR_GOTO(this->name, conf, out);
+
+ /* This builds volfile for volume level dameons */
+ glusterd_volume_svc_build_volfile_path(svc_name, volinfo, orgvol,
+ sizeof(orgvol));
+ /* Create the temporary volfile */
+ ret = gf_asprintf(&tmpvol, "/tmp/g%s-XXXXXX", svc_name);
+ if (ret < 0) {
+ goto out;
+ }
+
+ /* coverity[secure_temp] mkstemp uses 0600 as the mode and is safe */
+ tmpfd = mkstemp(tmpvol);
+ if (tmpfd < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, errno, GD_MSG_FILE_OP_FAILED,
+ "Unable to create temp file"
+ " %s:(%s)",
+ tmpvol, strerror(errno));
+ ret = -1;
+ goto out;
+ }
+
+ tmpclean = 1; /* SET the flag to unlink() tmpfile */
+
+ ret = builder(volinfo, tmpvol, mode_dict);
+ if (ret)
+ goto out;
+
+ /* Compare the topology of volfiles */
+ ret = glusterd_check_topology_identical(orgvol, tmpvol, identical);
+out:
+ if (tmpfd >= 0)
+ sys_close(tmpfd);
+ if (tmpclean)
+ sys_unlink(tmpvol);
+ if (tmpvol != NULL)
+ GF_FREE(tmpvol);
+ return ret;
+}
+
+void *
+__gf_find_compatible_svc(gd_node_type daemon)
+{
+ glusterd_svc_proc_t *svc_proc = NULL;
+ glusterd_svc_proc_t *return_proc = NULL;
+ glusterd_svc_t *parent_svc = NULL;
+ struct cds_list_head *svc_procs = NULL;
+ glusterd_conf_t *conf = NULL;
+ int pid = -1;
+
+ conf = THIS->private;
+ GF_VALIDATE_OR_GOTO("glusterd", conf, out);
+
+ if (daemon == GD_NODE_SHD) {
+ svc_procs = &conf->shd_procs;
+ if (!svc_procs)
+ goto out;
+ }
+
+ cds_list_for_each_entry(svc_proc, svc_procs, svc_proc_list)
+ {
+ parent_svc = cds_list_entry(svc_proc->svcs.next, glusterd_svc_t,
+ mux_svc);
+ if (!return_proc)
+ return_proc = svc_proc;
+
+ /* If there is an already running shd daemons, select it. Otehrwise
+ * select the first one.
+ */
+ if (parent_svc && gf_is_service_running(parent_svc->proc.pidfile, &pid))
+ return (void *)svc_proc;
+ /*
+ * Logic to select one process goes here. Currently there is only one
+ * shd_proc. So selecting the first one;
+ */
+ }
+out:
+ return return_proc;
+}
+
+glusterd_svc_proc_t *
+glusterd_svcprocess_new()
+{
+ glusterd_svc_proc_t *new_svcprocess = NULL;
+
+ new_svcprocess = GF_CALLOC(1, sizeof(*new_svcprocess),
+ gf_gld_mt_glusterd_svc_proc_t);
+
+ if (!new_svcprocess)
+ return NULL;
+
+ CDS_INIT_LIST_HEAD(&new_svcprocess->svc_proc_list);
+ CDS_INIT_LIST_HEAD(&new_svcprocess->svcs);
+ new_svcprocess->notify = glusterd_muxsvc_common_rpc_notify;
+ return new_svcprocess;
+}
+
+int
+glusterd_shd_svc_mux_init(glusterd_volinfo_t *volinfo, glusterd_svc_t *svc)
+{
+ int ret = -1;
+ glusterd_svc_proc_t *mux_proc = NULL;
+ glusterd_conn_t *mux_conn = NULL;
+ glusterd_conf_t *conf = NULL;
+ glusterd_svc_t *parent_svc = NULL;
+ int pid = -1;
+
+ GF_VALIDATE_OR_GOTO("glusterd", svc, out);
+ GF_VALIDATE_OR_GOTO("glusterd", volinfo, out);
+ conf = THIS->private;
+ GF_VALIDATE_OR_GOTO("glusterd", conf, out);
+ GF_VALIDATE_OR_GOTO("glusterd", svc, out);
+
+ pthread_mutex_lock(&conf->attach_lock);
+ {
+ if (!svc->inited) {
+ if (gf_is_service_running(svc->proc.pidfile, &pid)) {
+ /* Just connect is required, but we don't know what happens
+ * during the disconnect. So better to reattach.
+ */
+ mux_proc = __gf_find_compatible_svc_from_pid(GD_NODE_SHD, pid);
+ }
+
+ if (!mux_proc) {
+ if (pid != -1 && sys_access(svc->proc.pidfile, R_OK) == 0) {
+ /* stale pid file, unlink it. */
+ kill(pid, SIGTERM);
+ sys_unlink(svc->proc.pidfile);
+ }
+ mux_proc = __gf_find_compatible_svc(GD_NODE_SHD);
+ }
+ if (mux_proc) {
+ /* Take first entry from the process */
+ parent_svc = cds_list_entry(mux_proc->svcs.next, glusterd_svc_t,
+ mux_svc);
+ sys_link(parent_svc->proc.pidfile, svc->proc.pidfile);
+ mux_conn = &parent_svc->conn;
+ if (volinfo)
+ volinfo->shd.attached = _gf_true;
+ } else {
+ mux_proc = glusterd_svcprocess_new();
+ if (!mux_proc) {
+ ret = -1;
+ goto unlock;
+ }
+ cds_list_add_tail(&mux_proc->svc_proc_list, &conf->shd_procs);
+ }
+ svc->svc_proc = mux_proc;
+ cds_list_del_init(&svc->mux_svc);
+ cds_list_add_tail(&svc->mux_svc, &mux_proc->svcs);
+ ret = glusterd_shdsvc_init(volinfo, mux_conn, mux_proc);
+ if (ret) {
+ pthread_mutex_unlock(&conf->attach_lock);
+ gf_msg(THIS->name, GF_LOG_ERROR, 0, GD_MSG_FAILED_INIT_SHDSVC,
+ "Failed to init shd "
+ "service");
+ goto out;
+ }
+ gf_msg_debug(THIS->name, 0, "shd service initialized");
+ svc->inited = _gf_true;
+ }
+ ret = 0;
+ }
+unlock:
+ pthread_mutex_unlock(&conf->attach_lock);
+out:
+ return ret;
+}
+
+void *
+__gf_find_compatible_svc_from_pid(gd_node_type daemon, pid_t pid)
+{
+ glusterd_svc_proc_t *svc_proc = NULL;
+ struct cds_list_head *svc_procs = NULL;
+ glusterd_svc_t *svc = NULL;
+ pid_t mux_pid = -1;
+ glusterd_conf_t *conf = NULL;
+
+ conf = THIS->private;
+ if (!conf)
+ return NULL;
+
+ if (daemon == GD_NODE_SHD) {
+ svc_procs = &conf->shd_procs;
+ if (!svc_proc)
+ return NULL;
+ } /* Can be moved to switch when mux is implemented for other daemon; */
+
+ cds_list_for_each_entry(svc_proc, svc_procs, svc_proc_list)
+ {
+ cds_list_for_each_entry(svc, &svc_proc->svcs, mux_svc)
+ {
+ if (gf_is_service_running(svc->proc.pidfile, &mux_pid)) {
+ if (mux_pid == pid) {
+ /*TODO
+ * inefficient loop, but at the moment, there is only
+ * one shd.
+ */
+ return svc_proc;
+ }
+ }
+ }
+ }
+ return NULL;
+}
+
+static int32_t
+my_callback(struct rpc_req *req, struct iovec *iov, int count, void *v_frame)
+{
+ call_frame_t *frame = v_frame;
+ xlator_t *this = NULL;
+ glusterd_conf_t *conf = NULL;
+
+ GF_VALIDATE_OR_GOTO("glusterd", frame, out);
+ this = frame->this;
+ GF_VALIDATE_OR_GOTO("glusterd", this, out);
+ conf = this->private;
+ GF_VALIDATE_OR_GOTO(this->name, conf, out);
+
+ GF_ATOMIC_DEC(conf->blockers);
+
+ STACK_DESTROY(frame->root);
+out:
+ return 0;
+}
+
+static int32_t
+glusterd_svc_attach_cbk(struct rpc_req *req, struct iovec *iov, int count,
+ void *v_frame)
+{
+ call_frame_t *frame = v_frame;
+ glusterd_volinfo_t *volinfo = NULL;
+ glusterd_shdsvc_t *shd = NULL;
+ glusterd_svc_t *svc = frame->cookie;
+ glusterd_svc_t *parent_svc = NULL;
+ glusterd_svc_proc_t *mux_proc = NULL;
+ glusterd_conf_t *conf = NULL;
+ int *flag = (int *)frame->local;
+ xlator_t *this = THIS;
+ int pid = -1;
+ int ret = -1;
+ gf_getspec_rsp rsp = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("glusterd", this, out);
+ conf = this->private;
+ GF_VALIDATE_OR_GOTO("glusterd", conf, out);
+ GF_VALIDATE_OR_GOTO("glusterd", frame, out);
+ GF_VALIDATE_OR_GOTO("glusterd", svc, out);
+
+ frame->local = NULL;
+ frame->cookie = NULL;
+
+ if (!strcmp(svc->name, "glustershd")) {
+ /* Get volinfo->shd from svc object */
+ shd = cds_list_entry(svc, glusterd_shdsvc_t, svc);
+ if (!shd) {
+ gf_msg("glusterd", GF_LOG_ERROR, 0, GD_MSG_SHD_OBJ_GET_FAIL,
+ "Failed to get shd object "
+ "from shd service");
+ goto out;
+ }
+
+ /* Get volinfo from shd */
+ volinfo = cds_list_entry(shd, glusterd_volinfo_t, shd);
+ if (!volinfo) {
+ gf_msg("glusterd", GF_LOG_ERROR, 0, GD_MSG_VOLINFO_GET_FAIL,
+ "Failed to get volinfo from "
+ "from shd");
+ goto out;
+ }
+ }
+
+ if (!iov) {
+ gf_msg(frame->this->name, GF_LOG_ERROR, 0, GD_MSG_REQ_DECODE_FAIL,
+ "iov is NULL");
+ ret = -1;
+ goto out;
+ }
+
+ ret = xdr_to_generic(*iov, &rsp, (xdrproc_t)xdr_gf_getspec_rsp);
+ if (ret < 0) {
+ gf_msg(frame->this->name, GF_LOG_ERROR, 0, GD_MSG_REQ_DECODE_FAIL,
+ "XDR decoding error");
+ ret = -1;
+ goto out;
+ }
+
+ if (rsp.op_ret == 0) {
+ pthread_mutex_lock(&conf->attach_lock);
+ {
+ if (!strcmp(svc->name, "glustershd")) {
+ mux_proc = svc->svc_proc;
+ if (mux_proc &&
+ !gf_is_service_running(svc->proc.pidfile, &pid)) {
+ /*
+ * When svc's are restarting, there is a chance that the
+ * attached svc might not have updated it's pid. Because
+ * it was at connection stage. So in that case, we need
+ * to retry the pid file copy.
+ */
+ parent_svc = cds_list_entry(mux_proc->svcs.next,
+ glusterd_svc_t, mux_svc);
+ if (parent_svc)
+ sys_link(parent_svc->proc.pidfile, svc->proc.pidfile);
+ }
+ }
+ svc->online = _gf_true;
+ }
+ pthread_mutex_unlock(&conf->attach_lock);
+ gf_msg(this->name, GF_LOG_INFO, 0, GD_MSG_SVC_ATTACH_FAIL,
+ "svc %s of volume %s attached successfully to pid %d", svc->name,
+ volinfo->volname, glusterd_proc_get_pid(&svc->proc));
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_SVC_ATTACH_FAIL,
+ "svc %s of volume %s failed to "
+ "attach to pid %d. Starting a new process",
+ svc->name, volinfo->volname, glusterd_proc_get_pid(&svc->proc));
+ if (!strcmp(svc->name, "glustershd")) {
+ glusterd_recover_shd_attach_failure(volinfo, svc, *flag);
+ }
+ }
+out:
+ if (flag) {
+ GF_FREE(flag);
+ }
+ GF_ATOMIC_DEC(conf->blockers);
+ STACK_DESTROY(frame->root);
+ return 0;
+}
+
+extern size_t
+build_volfile_path(char *volume_id, char *path, size_t path_len,
+ char *trusted_str);
+
+int
+__glusterd_send_svc_configure_req(glusterd_svc_t *svc, int flags,
+ struct rpc_clnt *rpc, char *volfile_id,
+ int op)
+{
+ int ret = -1;
+ struct iobuf *iobuf = NULL;
+ struct iobref *iobref = NULL;
+ struct iovec iov = {
+ 0,
+ };
+ char path[PATH_MAX] = {
+ '\0',
+ };
+ struct stat stbuf = {
+ 0,
+ };
+ int32_t spec_fd = -1;
+ size_t file_len = -1;
+ char *volfile_content = NULL;
+ ssize_t req_size = 0;
+ call_frame_t *frame = NULL;
+ gd1_mgmt_brick_op_req brick_req;
+ void *req = &brick_req;
+ void *errlbl = &&err;
+ struct rpc_clnt_connection *conn;
+ xlator_t *this = THIS;
+ glusterd_conf_t *conf = THIS->private;
+ extern struct rpc_clnt_program gd_brick_prog;
+ fop_cbk_fn_t cbkfn = my_callback;
+
+ if (!rpc) {
+ gf_msg("glusterd", GF_LOG_ERROR, 0, GD_MSG_PARAM_NULL,
+ "called with null rpc");
+ return -1;
+ }
+
+ conn = &rpc->conn;
+ if (!conn->connected || conn->disconnected) {
+ gf_msg(this->name, GF_LOG_INFO, 0, GD_MSG_CONNECT_RETURNED,
+ "not connected yet");
+ return -1;
+ }
+
+ brick_req.op = op;
+ brick_req.name = volfile_id;
+ brick_req.input.input_val = NULL;
+ brick_req.input.input_len = 0;
+
+ frame = create_frame(this, this->ctx->pool);
+ if (!frame) {
+ goto *errlbl;
+ }
+
+ if (op == GLUSTERD_SVC_ATTACH) {
+ (void)build_volfile_path(volfile_id, path, sizeof(path), NULL);
+
+ ret = sys_stat(path, &stbuf);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_SVC_ATTACH_FAIL,
+ "Unable to stat %s (%s)", path, strerror(errno));
+ ret = -EINVAL;
+ goto *errlbl;
+ }
+
+ file_len = stbuf.st_size;
+ volfile_content = GF_MALLOC(file_len + 1, gf_common_mt_char);
+ if (!volfile_content) {
+ ret = -ENOMEM;
+ goto *errlbl;
+ }
+ spec_fd = open(path, O_RDONLY);
+ if (spec_fd < 0) {
+ gf_msg(THIS->name, GF_LOG_WARNING, 0, GD_MSG_SVC_ATTACH_FAIL,
+ "failed to read volfile %s", path);
+ ret = -EIO;
+ goto *errlbl;
+ }
+ ret = sys_read(spec_fd, volfile_content, file_len);
+ if (ret == file_len) {
+ brick_req.input.input_val = volfile_content;
+ brick_req.input.input_len = file_len;
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_SVC_ATTACH_FAIL,
+ "read failed on path %s. File size=%" GF_PRI_SIZET
+ "read size=%d",
+ path, file_len, ret);
+ ret = -EIO;
+ goto *errlbl;
+ }
+
+ frame->cookie = svc;
+ frame->local = GF_CALLOC(1, sizeof(int), gf_gld_mt_int);
+ *((int *)frame->local) = flags;
+ cbkfn = glusterd_svc_attach_cbk;
+ }
+
+ req_size = xdr_sizeof((xdrproc_t)xdr_gd1_mgmt_brick_op_req, req);
+ iobuf = iobuf_get2(rpc->ctx->iobuf_pool, req_size);
+ if (!iobuf) {
+ goto *errlbl;
+ }
+ errlbl = &&maybe_free_iobuf;
+
+ iov.iov_base = iobuf->ptr;
+ iov.iov_len = iobuf_pagesize(iobuf);
+
+ iobref = iobref_new();
+ if (!iobref) {
+ goto *errlbl;
+ }
+ errlbl = &&free_iobref;
+
+ iobref_add(iobref, iobuf);
+ /*
+ * Drop our reference to the iobuf. The iobref should already have
+ * one after iobref_add, so when we unref that we'll free the iobuf as
+ * well. This allows us to pass just the iobref as frame->local.
+ */
+ iobuf_unref(iobuf);
+ /* Set the pointer to null so we don't free it on a later error. */
+ iobuf = NULL;
+
+ /* Create the xdr payload */
+ ret = xdr_serialize_generic(iov, req, (xdrproc_t)xdr_gd1_mgmt_brick_op_req);
+ if (ret == -1) {
+ goto *errlbl;
+ }
+ iov.iov_len = ret;
+
+ /* Send the msg */
+ GF_ATOMIC_INC(conf->blockers);
+ ret = rpc_clnt_submit(rpc, &gd_brick_prog, op, cbkfn, &iov, 1, NULL, 0,
+ iobref, frame, NULL, 0, NULL, 0, NULL);
+ GF_FREE(volfile_content);
+ if (spec_fd >= 0)
+ sys_close(spec_fd);
+ return ret;
+
+free_iobref:
+ iobref_unref(iobref);
+maybe_free_iobuf:
+ if (iobuf) {
+ iobuf_unref(iobuf);
+ }
+err:
+ GF_FREE(volfile_content);
+ if (spec_fd >= 0)
+ sys_close(spec_fd);
+ if (frame)
+ STACK_DESTROY(frame->root);
+ return -1;
+}
+
+int
+glusterd_attach_svc(glusterd_svc_t *svc, glusterd_volinfo_t *volinfo, int flags)
+{
+ glusterd_conf_t *conf = THIS->private;
+ int ret = -1;
+ int tries;
+ rpc_clnt_t *rpc = NULL;
+
+ GF_VALIDATE_OR_GOTO("glusterd", conf, out);
+ GF_VALIDATE_OR_GOTO("glusterd", svc, out);
+ GF_VALIDATE_OR_GOTO("glusterd", volinfo, out);
+
+ gf_msg("glusterd", GF_LOG_INFO, 0, GD_MSG_ATTACH_INFO,
+ "adding svc %s (volume=%s) to existing "
+ "process with pid %d",
+ svc->name, volinfo->volname, glusterd_proc_get_pid(&svc->proc));
+
+ rpc = rpc_clnt_ref(svc->conn.rpc);
+ for (tries = 15; tries > 0; --tries) {
+ if (rpc) {
+ pthread_mutex_lock(&conf->attach_lock);
+ {
+ ret = __glusterd_send_svc_configure_req(
+ svc, flags, rpc, svc->proc.volfileid, GLUSTERD_SVC_ATTACH);
+ }
+ pthread_mutex_unlock(&conf->attach_lock);
+ if (!ret) {
+ volinfo->shd.attached = _gf_true;
+ goto out;
+ }
+ }
+ /*
+ * It might not actually be safe to manipulate the lock
+ * like this, but if we don't then the connection can
+ * never actually complete and retries are useless.
+ * Unfortunately, all of the alternatives (e.g. doing
+ * all of this in a separate thread) are much more
+ * complicated and risky.
+ * TBD: see if there's a better way
+ */
+ synclock_unlock(&conf->big_lock);
+ sleep(1);
+ synclock_lock(&conf->big_lock);
+ }
+ ret = -1;
+ gf_msg("glusterd", GF_LOG_WARNING, 0, GD_MSG_SVC_ATTACH_FAIL,
+ "attach failed for %s(volume=%s)", svc->name, volinfo->volname);
+out:
+ if (rpc)
+ rpc_clnt_unref(rpc);
+ return ret;
+}
+
+int
+glusterd_detach_svc(glusterd_svc_t *svc, glusterd_volinfo_t *volinfo, int sig)
+{
+ glusterd_conf_t *conf = THIS->private;
+ int ret = -1;
+ int tries;
+ rpc_clnt_t *rpc = NULL;
+
+ GF_VALIDATE_OR_GOTO(THIS->name, conf, out);
+ GF_VALIDATE_OR_GOTO(THIS->name, svc, out);
+ GF_VALIDATE_OR_GOTO(THIS->name, volinfo, out);
+
+ gf_msg(THIS->name, GF_LOG_INFO, 0, GD_MSG_DETACH_INFO,
+ "removing svc %s (volume=%s) from existing "
+ "process with pid %d",
+ svc->name, volinfo->volname, glusterd_proc_get_pid(&svc->proc));
+
+ rpc = rpc_clnt_ref(svc->conn.rpc);
+ for (tries = 15; tries > 0; --tries) {
+ if (rpc) {
+ /*For detach there is no flags, and we are not using sig.*/
+ pthread_mutex_lock(&conf->attach_lock);
+ {
+ ret = __glusterd_send_svc_configure_req(svc, 0, svc->conn.rpc,
+ svc->proc.volfileid,
+ GLUSTERD_SVC_DETACH);
+ }
+ pthread_mutex_unlock(&conf->attach_lock);
+ if (!ret) {
+ goto out;
+ }
+ }
+ /*
+ * It might not actually be safe to manipulate the lock
+ * like this, but if we don't then the connection can
+ * never actually complete and retries are useless.
+ * Unfortunately, all of the alternatives (e.g. doing
+ * all of this in a separate thread) are much more
+ * complicated and risky.
+ * TBD: see if there's a better way
+ */
+ synclock_unlock(&conf->big_lock);
+ sleep(1);
+ synclock_lock(&conf->big_lock);
+ }
+ ret = -1;
+ gf_msg("glusterd", GF_LOG_WARNING, 0, GD_MSG_SVC_DETACH_FAIL,
+ "detach failed for %s(volume=%s)", svc->name, volinfo->volname);
+out:
+ if (rpc)
+ rpc_clnt_unref(rpc);
+ return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-svc-helper.h b/xlators/mgmt/glusterd/src/glusterd-svc-helper.h
index cc98e78..5def246 100644
--- a/xlators/mgmt/glusterd/src/glusterd-svc-helper.h
+++ b/xlators/mgmt/glusterd/src/glusterd-svc-helper.h
@@ -16,10 +16,10 @@
#include "glusterd-volgen.h"
int
-glusterd_svcs_reconfigure();
+glusterd_svcs_reconfigure(glusterd_volinfo_t *volinfo);
int
-glusterd_svcs_stop();
+glusterd_svcs_stop(glusterd_volinfo_t *vol);
int
glusterd_svcs_manager(glusterd_volinfo_t *volinfo);
@@ -41,5 +41,41 @@ int
glusterd_svc_check_tier_topology_identical(char *svc_name,
glusterd_volinfo_t *volinfo,
gf_boolean_t *identical);
+int
+glusterd_volume_svc_check_volfile_identical(char *svc_name, dict_t *mode_dict,
+ glusterd_volinfo_t *volinfo,
+ glusterd_vol_graph_builder_t,
+ gf_boolean_t *identical);
+int
+glusterd_volume_svc_check_topology_identical(char *svc_name, dict_t *mode_dict,
+ glusterd_volinfo_t *volinfo,
+ glusterd_vol_graph_builder_t,
+ gf_boolean_t *identical);
+void
+glusterd_volume_svc_build_volfile_path(char *server, glusterd_volinfo_t *vol,
+ char *volfile, size_t len);
+void *
+__gf_find_compatible_svc(gd_node_type daemon);
+
+glusterd_svc_proc_t *
+glusterd_svcprocess_new();
+
+int
+glusterd_shd_svc_mux_init(glusterd_volinfo_t *volinfo, glusterd_svc_t *svc);
+
+void *
+__gf_find_compatible_svc_from_pid(gd_node_type daemon, pid_t pid);
+
+int
+glusterd_attach_svc(glusterd_svc_t *svc, glusterd_volinfo_t *volinfo,
+ int flags);
+
+int
+glusterd_detach_svc(glusterd_svc_t *svc, glusterd_volinfo_t *volinfo, int sig);
+
+int
+__glusterd_send_svc_configure_req(glusterd_svc_t *svc, int flag,
+ struct rpc_clnt *rpc, char *volfile_id,
+ int op);
#endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-svc-mgmt.c b/xlators/mgmt/glusterd/src/glusterd-svc-mgmt.c
index 4cd4cea..f32dafc 100644
--- a/xlators/mgmt/glusterd/src/glusterd-svc-mgmt.c
+++ b/xlators/mgmt/glusterd/src/glusterd-svc-mgmt.c
@@ -18,6 +18,7 @@
#include "glusterd-conn-mgmt.h"
#include "glusterd-messages.h"
#include <glusterfs/syscall.h>
+#include "glusterd-shd-svc-helper.h"
int
glusterd_svc_create_rundir(char *rundir)
@@ -167,68 +168,75 @@ glusterd_svc_start(glusterd_svc_t *svc, int flags, dict_t *cmdline)
GF_ASSERT(this);
priv = this->private;
- GF_ASSERT(priv);
+ GF_VALIDATE_OR_GOTO("glusterd", priv, out);
+ GF_VALIDATE_OR_GOTO("glusterd", svc, out);
+
+ pthread_mutex_lock(&priv->attach_lock);
+ {
+ if (glusterd_proc_is_running(&(svc->proc))) {
+ ret = 0;
+ goto unlock;
+ }
- if (glusterd_proc_is_running(&(svc->proc))) {
- ret = 0;
- goto out;
- }
+ ret = sys_access(svc->proc.volfile, F_OK);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_VOLFILE_NOT_FOUND,
+ "Volfile %s is not present", svc->proc.volfile);
+ goto unlock;
+ }
- ret = sys_access(svc->proc.volfile, F_OK);
- if (ret) {
- gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_VOLFILE_NOT_FOUND,
- "Volfile %s is not present", svc->proc.volfile);
- goto out;
- }
+ runinit(&runner);
- runinit(&runner);
+ if (this->ctx->cmd_args.valgrind) {
+ len = snprintf(valgrind_logfile, PATH_MAX, "%s/valgrind-%s.log",
+ svc->proc.logfile, svc->name);
+ if ((len < 0) || (len >= PATH_MAX)) {
+ ret = -1;
+ goto unlock;
+ }
- if (this->ctx->cmd_args.valgrind) {
- len = snprintf(valgrind_logfile, PATH_MAX, "%s/valgrind-%s.log",
- svc->proc.logfile, svc->name);
- if ((len < 0) || (len >= PATH_MAX)) {
- ret = -1;
- goto out;
+ runner_add_args(&runner, "valgrind", "--leak-check=full",
+ "--trace-children=yes", "--track-origins=yes",
+ NULL);
+ runner_argprintf(&runner, "--log-file=%s", valgrind_logfile);
}
- runner_add_args(&runner, "valgrind", "--leak-check=full",
- "--trace-children=yes", "--track-origins=yes", NULL);
- runner_argprintf(&runner, "--log-file=%s", valgrind_logfile);
- }
-
- runner_add_args(&runner, SBIN_DIR "/glusterfs", "-s",
- svc->proc.volfileserver, "--volfile-id",
- svc->proc.volfileid, "-p", svc->proc.pidfile, "-l",
- svc->proc.logfile, "-S", svc->conn.sockpath, NULL);
+ runner_add_args(&runner, SBIN_DIR "/glusterfs", "-s",
+ svc->proc.volfileserver, "--volfile-id",
+ svc->proc.volfileid, "-p", svc->proc.pidfile, "-l",
+ svc->proc.logfile, "-S", svc->conn.sockpath, NULL);
- if (dict_get_strn(priv->opts, GLUSTERD_LOCALTIME_LOGGING_KEY,
- SLEN(GLUSTERD_LOCALTIME_LOGGING_KEY),
- &localtime_logging) == 0) {
- if (strcmp(localtime_logging, "enable") == 0)
- runner_add_arg(&runner, "--localtime-logging");
- }
- if (dict_get_strn(priv->opts, GLUSTERD_DAEMON_LOG_LEVEL_KEY,
- SLEN(GLUSTERD_DAEMON_LOG_LEVEL_KEY), &log_level) == 0) {
- snprintf(daemon_log_level, 30, "--log-level=%s", log_level);
- runner_add_arg(&runner, daemon_log_level);
- }
+ if (dict_get_strn(priv->opts, GLUSTERD_LOCALTIME_LOGGING_KEY,
+ SLEN(GLUSTERD_LOCALTIME_LOGGING_KEY),
+ &localtime_logging) == 0) {
+ if (strcmp(localtime_logging, "enable") == 0)
+ runner_add_arg(&runner, "--localtime-logging");
+ }
+ if (dict_get_strn(priv->opts, GLUSTERD_DAEMON_LOG_LEVEL_KEY,
+ SLEN(GLUSTERD_DAEMON_LOG_LEVEL_KEY),
+ &log_level) == 0) {
+ snprintf(daemon_log_level, 30, "--log-level=%s", log_level);
+ runner_add_arg(&runner, daemon_log_level);
+ }
- if (cmdline)
- dict_foreach(cmdline, svc_add_args, (void *)&runner);
+ if (cmdline)
+ dict_foreach(cmdline, svc_add_args, (void *)&runner);
- gf_msg(this->name, GF_LOG_INFO, 0, GD_MSG_SVC_START_SUCCESS,
- "Starting %s service", svc->name);
+ gf_msg(this->name, GF_LOG_INFO, 0, GD_MSG_SVC_START_SUCCESS,
+ "Starting %s service", svc->name);
- if (flags == PROC_START_NO_WAIT) {
- ret = runner_run_nowait(&runner);
- } else {
- synclock_unlock(&priv->big_lock);
- {
- ret = runner_run(&runner);
+ if (flags == PROC_START_NO_WAIT) {
+ ret = runner_run_nowait(&runner);
+ } else {
+ synclock_unlock(&priv->big_lock);
+ {
+ ret = runner_run(&runner);
+ }
+ synclock_lock(&priv->big_lock);
}
- synclock_lock(&priv->big_lock);
}
-
+unlock:
+ pthread_mutex_unlock(&priv->attach_lock);
out:
gf_msg_debug(this->name, 0, "Returning %d", ret);
@@ -281,7 +289,8 @@ glusterd_svc_build_volfile_path(char *server, char *workdir, char *volfile,
glusterd_svc_build_svcdir(server, workdir, dir, sizeof(dir));
- if (!strcmp(server, "quotad")) /*quotad has different volfile name*/
+ if (!strcmp(server, "quotad"))
+ /*quotad has different volfile name*/
snprintf(volfile, len, "%s/%s.vol", dir, server);
else
snprintf(volfile, len, "%s/%s-server.vol", dir, server);
@@ -366,3 +375,138 @@ glusterd_svc_common_rpc_notify(glusterd_conn_t *conn, rpc_clnt_event_t event)
return ret;
}
+
+void
+glusterd_volume_svc_build_volfile_path(char *server, glusterd_volinfo_t *vol,
+ char *volfile, size_t len)
+{
+ GF_ASSERT(len == PATH_MAX);
+
+ if (!strcmp(server, "glustershd")) {
+ glusterd_svc_build_shd_volfile_path(vol, volfile, len);
+ }
+}
+
+int
+glusterd_muxsvc_common_rpc_notify(glusterd_svc_proc_t *mux_proc,
+ rpc_clnt_event_t event)
+{
+ int ret = 0;
+ glusterd_svc_t *svc = NULL;
+ glusterd_svc_t *tmp = NULL;
+ xlator_t *this = NULL;
+ gf_boolean_t need_logging = _gf_false;
+
+ this = THIS;
+ GF_ASSERT(this);
+
+ if (!mux_proc) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_SVC_GET_FAIL,
+ "Failed to get the svc proc data");
+ return -1;
+ }
+
+ /* Currently this function was used for shd svc, if this function is
+ * using for another svc, change ths glustershd reference. We can get
+ * the svc name from any of the attached svc's
+ */
+ switch (event) {
+ case RPC_CLNT_CONNECT:
+ gf_msg_debug(this->name, 0,
+ "glustershd has connected with glusterd.");
+ gf_event(EVENT_SVC_CONNECTED, "svc_name=glustershd");
+ cds_list_for_each_entry_safe(svc, tmp, &mux_proc->svcs, mux_svc)
+ {
+ if (svc->online)
+ continue;
+ svc->online = _gf_true;
+ }
+ break;
+
+ case RPC_CLNT_DISCONNECT:
+ cds_list_for_each_entry_safe(svc, tmp, &mux_proc->svcs, mux_svc)
+ {
+ if (svc->online) {
+ if (!need_logging)
+ need_logging = _gf_true;
+ svc->online = _gf_false;
+ }
+ }
+ if (need_logging) {
+ gf_msg(this->name, GF_LOG_INFO, 0, GD_MSG_NODE_DISCONNECTED,
+ "glustershd has disconnected from glusterd.");
+ gf_event(EVENT_SVC_DISCONNECTED, "svc_name=glustershd");
+ }
+ break;
+
+ default:
+ gf_msg_trace(this->name, 0, "got some other RPC event %d", event);
+ break;
+ }
+
+ return ret;
+}
+
+int
+glusterd_muxsvc_conn_init(glusterd_conn_t *conn, glusterd_svc_proc_t *mux_proc,
+ char *sockpath, int frame_timeout,
+ glusterd_muxsvc_conn_notify_t notify)
+{
+ int ret = -1;
+ dict_t *options = NULL;
+ struct rpc_clnt *rpc = NULL;
+ xlator_t *this = THIS;
+ glusterd_svc_t *svc = NULL;
+
+ options = dict_new();
+ if (!this || !options)
+ goto out;
+
+ svc = cds_list_entry(conn, glusterd_svc_t, conn);
+ if (!svc) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_SVC_GET_FAIL,
+ "Failed to get the service");
+ goto out;
+ }
+
+ ret = rpc_transport_unix_options_build(options, sockpath, frame_timeout);
+ if (ret)
+ goto out;
+
+ ret = dict_set_int32n(options, "transport.socket.ignore-enoent",
+ SLEN("transport.socket.ignore-enoent"), 1);
+ if (ret)
+ goto out;
+
+ /* @options is free'd by rpc_transport when destroyed */
+ rpc = rpc_clnt_new(options, this, (char *)svc->name, 16);
+ if (!rpc) {
+ ret = -1;
+ goto out;
+ }
+
+ ret = rpc_clnt_register_notify(rpc, glusterd_muxsvc_conn_common_notify,
+ mux_proc);
+ if (ret)
+ goto out;
+
+ ret = snprintf(conn->sockpath, sizeof(conn->sockpath), "%s", sockpath);
+ if (ret < 0)
+ goto out;
+ else
+ ret = 0;
+
+ conn->frame_timeout = frame_timeout;
+ conn->rpc = rpc;
+ mux_proc->notify = notify;
+out:
+ if (options)
+ dict_unref(options);
+ if (ret) {
+ if (rpc) {
+ rpc_clnt_unref(rpc);
+ rpc = NULL;
+ }
+ }
+ return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-svc-mgmt.h b/xlators/mgmt/glusterd/src/glusterd-svc-mgmt.h
index c850bfd..fbc5225 100644
--- a/xlators/mgmt/glusterd/src/glusterd-svc-mgmt.h
+++ b/xlators/mgmt/glusterd/src/glusterd-svc-mgmt.h
@@ -13,9 +13,12 @@
#include "glusterd-proc-mgmt.h"
#include "glusterd-conn-mgmt.h"
+#include "glusterd-rcu.h"
struct glusterd_svc_;
+
typedef struct glusterd_svc_ glusterd_svc_t;
+typedef struct glusterd_svc_proc_ glusterd_svc_proc_t;
typedef void (*glusterd_svc_build_t)(glusterd_svc_t *svc);
@@ -25,6 +28,17 @@ typedef int (*glusterd_svc_start_t)(glusterd_svc_t *svc, int flags);
typedef int (*glusterd_svc_stop_t)(glusterd_svc_t *svc, int sig);
typedef int (*glusterd_svc_reconfigure_t)(void *data);
+typedef int (*glusterd_muxsvc_conn_notify_t)(glusterd_svc_proc_t *mux_proc,
+ rpc_clnt_event_t event);
+
+struct glusterd_svc_proc_ {
+ struct cds_list_head svc_proc_list;
+ struct cds_list_head svcs;
+ glusterd_muxsvc_conn_notify_t notify;
+ rpc_clnt_t *rpc;
+ void *data;
+};
+
struct glusterd_svc_ {
char name[NAME_MAX];
glusterd_conn_t conn;
@@ -35,6 +49,8 @@ struct glusterd_svc_ {
gf_boolean_t online;
gf_boolean_t inited;
glusterd_svc_reconfigure_t reconfigure;
+ glusterd_svc_proc_t *svc_proc;
+ struct cds_list_head mux_svc;
};
int
@@ -69,4 +85,15 @@ glusterd_svc_reconfigure(int (*create_volfile)());
int
glusterd_svc_common_rpc_notify(glusterd_conn_t *conn, rpc_clnt_event_t event);
+int
+glusterd_muxsvc_common_rpc_notify(glusterd_svc_proc_t *conn,
+ rpc_clnt_event_t event);
+
+int
+glusterd_proc_get_pid(glusterd_proc_t *proc);
+
+int
+glusterd_muxsvc_conn_init(glusterd_conn_t *conn, glusterd_svc_proc_t *mux_proc,
+ char *sockpath, int frame_timeout,
+ glusterd_muxsvc_conn_notify_t notify);
#endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-tier.c b/xlators/mgmt/glusterd/src/glusterd-tier.c
index 4dc0d44..23a9592 100644
--- a/xlators/mgmt/glusterd/src/glusterd-tier.c
+++ b/xlators/mgmt/glusterd/src/glusterd-tier.c
@@ -27,6 +27,7 @@
#include "glusterd-messages.h"
#include "glusterd-mgmt.h"
#include "glusterd-syncop.h"
+#include "glusterd-shd-svc-helper.h"
#include <sys/wait.h>
#include <dlfcn.h>
@@ -615,7 +616,7 @@ glusterd_op_remove_tier_brick(dict_t *dict, char **op_errstr, dict_t *rsp_dict)
if (cmd == GF_DEFRAG_CMD_DETACH_START &&
volinfo->status == GLUSTERD_STATUS_STARTED) {
- ret = glusterd_svcs_reconfigure();
+ ret = glusterd_svcs_reconfigure(volinfo);
if (ret) {
gf_msg(this->name, GF_LOG_WARNING, 0, GD_MSG_NFS_RECONF_FAIL,
"Unable to reconfigure NFS-Server");
diff --git a/xlators/mgmt/glusterd/src/glusterd-tierd-svc.c b/xlators/mgmt/glusterd/src/glusterd-tierd-svc.c
index 04ceec5..ab463f1 100644
--- a/xlators/mgmt/glusterd/src/glusterd-tierd-svc.c
+++ b/xlators/mgmt/glusterd/src/glusterd-tierd-svc.c
@@ -83,7 +83,6 @@ glusterd_tierdsvc_init(void *data)
goto out;
notify = glusterd_svc_common_rpc_notify;
- glusterd_store_perform_node_state_store(volinfo);
volinfo->type = GF_CLUSTER_TYPE_TIER;
@@ -395,6 +394,7 @@ int
glusterd_tierdsvc_restart()
{
glusterd_volinfo_t *volinfo = NULL;
+ glusterd_volinfo_t *tmp = NULL;
int ret = 0;
xlator_t *this = THIS;
glusterd_conf_t *conf = NULL;
@@ -405,7 +405,7 @@ glusterd_tierdsvc_restart()
conf = this->private;
GF_VALIDATE_OR_GOTO(this->name, conf, out);
- cds_list_for_each_entry(volinfo, &conf->volumes, vol_list)
+ cds_list_for_each_entry_safe(volinfo, tmp, &conf->volumes, vol_list)
{
/* Start per volume tierd svc */
if (volinfo->status == GLUSTERD_STATUS_STARTED &&
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index 52b83ec..ef664c2 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -61,6 +61,7 @@
#include "glusterd-server-quorum.h"
#include <glusterfs/quota-common-utils.h>
#include <glusterfs/common-utils.h>
+#include "glusterd-shd-svc-helper.h"
#include "xdr-generic.h"
#include <sys/resource.h>
@@ -583,13 +584,17 @@ glusterd_volinfo_t *
glusterd_volinfo_unref(glusterd_volinfo_t *volinfo)
{
int refcnt = -1;
+ glusterd_conf_t *conf = THIS->private;
- pthread_mutex_lock(&volinfo->reflock);
+ pthread_mutex_lock(&conf->volume_lock);
{
- refcnt = --volinfo->refcnt;
+ pthread_mutex_lock(&volinfo->reflock);
+ {
+ refcnt = --volinfo->refcnt;
+ }
+ pthread_mutex_unlock(&volinfo->reflock);
}
- pthread_mutex_unlock(&volinfo->reflock);
-
+ pthread_mutex_unlock(&conf->volume_lock);
if (!refcnt) {
glusterd_volinfo_delete(volinfo);
return NULL;
@@ -661,6 +666,7 @@ glusterd_volinfo_new(glusterd_volinfo_t **volinfo)
glusterd_snapdsvc_build(&new_volinfo->snapd.svc);
glusterd_tierdsvc_build(&new_volinfo->tierd.svc);
glusterd_gfproxydsvc_build(&new_volinfo->gfproxyd.svc);
+ glusterd_shdsvc_build(&new_volinfo->shd.svc);
pthread_mutex_init(&new_volinfo->reflock, NULL);
*volinfo = glusterd_volinfo_ref(new_volinfo);
@@ -1026,11 +1032,11 @@ glusterd_volinfo_delete(glusterd_volinfo_t *volinfo)
gf_store_handle_destroy(volinfo->snapd.handle);
glusterd_auth_cleanup(volinfo);
+ glusterd_shd_svcproc_cleanup(&volinfo->shd);
pthread_mutex_destroy(&volinfo->reflock);
GF_FREE(volinfo);
ret = 0;
-
out:
gf_msg_debug(THIS->name, 0, "Returning %d", ret);
return ret;
@@ -3619,6 +3625,7 @@ glusterd_spawn_daemons(void *opaque)
ret = glusterd_snapdsvc_restart();
ret = glusterd_tierdsvc_restart();
ret = glusterd_gfproxydsvc_restart();
+ ret = glusterd_shdsvc_restart();
return ret;
}
@@ -4569,6 +4576,9 @@ glusterd_delete_stale_volume(glusterd_volinfo_t *stale_volinfo,
svc = &(stale_volinfo->snapd.svc);
(void)svc->manager(svc, stale_volinfo, PROC_START_NO_WAIT);
}
+ svc = &(stale_volinfo->shd.svc);
+ (void)svc->manager(svc, stale_volinfo, PROC_START_NO_WAIT);
+
(void)glusterd_volinfo_remove(stale_volinfo);
return 0;
@@ -4683,6 +4693,15 @@ glusterd_import_friend_volume(dict_t *peer_data, int count)
glusterd_volinfo_unref(old_volinfo);
}
+ ret = glusterd_store_volinfo(new_volinfo, GLUSTERD_VOLINFO_VER_AC_NONE);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_VOLINFO_STORE_FAIL,
+ "Failed to store "
+ "volinfo for volume %s",
+ new_volinfo->volname);
+ goto out;
+ }
+
if (glusterd_is_volume_started(new_volinfo)) {
(void)glusterd_start_bricks(new_volinfo);
if (glusterd_is_snapd_enabled(new_volinfo)) {
@@ -4691,15 +4710,10 @@ glusterd_import_friend_volume(dict_t *peer_data, int count)
gf_event(EVENT_SVC_MANAGER_FAILED, "svc_name=%s", svc->name);
}
}
- }
-
- ret = glusterd_store_volinfo(new_volinfo, GLUSTERD_VOLINFO_VER_AC_NONE);
- if (ret) {
- gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_VOLINFO_STORE_FAIL,
- "Failed to store "
- "volinfo for volume %s",
- new_volinfo->volname);
- goto out;
+ svc = &(new_volinfo->shd.svc);
+ if (svc->manager(svc, new_volinfo, PROC_START_NO_WAIT)) {
+ gf_event(EVENT_SVC_MANAGER_FAILED, "svc_name=%s", svc->name);
+ }
}
ret = glusterd_create_volfiles_and_notify_services(new_volinfo);
@@ -5174,9 +5188,7 @@ glusterd_add_node_to_dict(char *server, dict_t *dict, int count,
glusterd_svc_build_pidfile_path(server, priv->rundir, pidfile,
sizeof(pidfile));
- if (strcmp(server, priv->shd_svc.name) == 0)
- svc = &(priv->shd_svc);
- else if (strcmp(server, priv->nfs_svc.name) == 0)
+ if (strcmp(server, priv->nfs_svc.name) == 0)
svc = &(priv->nfs_svc);
else if (strcmp(server, priv->quotad_svc.name) == 0)
svc = &(priv->quotad_svc);
@@ -5207,9 +5219,6 @@ glusterd_add_node_to_dict(char *server, dict_t *dict, int count,
if (!strcmp(server, priv->nfs_svc.name))
ret = dict_set_nstrn(dict, key, keylen, "NFS Server",
SLEN("NFS Server"));
- else if (!strcmp(server, priv->shd_svc.name))
- ret = dict_set_nstrn(dict, key, keylen, "Self-heal Daemon",
- SLEN("Self-heal Daemon"));
else if (!strcmp(server, priv->quotad_svc.name))
ret = dict_set_nstrn(dict, key, keylen, "Quota Daemon",
SLEN("Quota Daemon"));
@@ -8773,6 +8782,21 @@ glusterd_friend_remove_cleanup_vols(uuid_t uuid)
"to stop snapd daemon service");
}
}
+
+ if (glusterd_is_shd_compatible_volume(volinfo)) {
+ /*
+ * Sending stop request for all volumes. So it is fine
+ * to send stop for mux shd
+ */
+ svc = &(volinfo->shd.svc);
+ ret = svc->stop(svc, SIGTERM);
+ if (ret) {
+ gf_msg(THIS->name, GF_LOG_ERROR, 0, GD_MSG_SVC_STOP_FAIL,
+ "Failed "
+ "to stop shd daemon service");
+ }
+ }
+
if (volinfo->type == GF_CLUSTER_TYPE_TIER) {
svc = &(volinfo->tierd.svc);
ret = svc->stop(svc, SIGTERM);
@@ -8798,7 +8822,7 @@ glusterd_friend_remove_cleanup_vols(uuid_t uuid)
}
/* Reconfigure all daemon services upon peer detach */
- ret = glusterd_svcs_reconfigure();
+ ret = glusterd_svcs_reconfigure(NULL);
if (ret) {
gf_msg(THIS->name, GF_LOG_ERROR, 0, GD_MSG_SVC_STOP_FAIL,
"Failed to reconfigure all daemon services.");
@@ -14350,3 +14374,74 @@ glusterd_is_profile_on(glusterd_volinfo_t *volinfo)
return _gf_true;
return _gf_false;
}
+
+int32_t
+glusterd_add_shd_to_dict(glusterd_volinfo_t *volinfo, dict_t *dict,
+ int32_t count)
+{
+ int ret = -1;
+ int32_t pid = -1;
+ int32_t brick_online = -1;
+ char key[64] = {0};
+ int keylen;
+ char *pidfile = NULL;
+ xlator_t *this = NULL;
+ char *uuid_str = NULL;
+
+ this = THIS;
+ GF_VALIDATE_OR_GOTO(THIS->name, this, out);
+
+ GF_VALIDATE_OR_GOTO(this->name, volinfo, out);
+ GF_VALIDATE_OR_GOTO(this->name, dict, out);
+
+ keylen = snprintf(key, sizeof(key), "brick%d.hostname", count);
+ ret = dict_set_nstrn(dict, key, keylen, "Self-heal Daemon",
+ SLEN("Self-heal Daemon"));
+ if (ret)
+ goto out;
+
+ keylen = snprintf(key, sizeof(key), "brick%d.path", count);
+ uuid_str = gf_strdup(uuid_utoa(MY_UUID));
+ if (!uuid_str) {
+ ret = -1;
+ goto out;
+ }
+ ret = dict_set_dynstrn(dict, key, keylen, uuid_str);
+ if (ret)
+ goto out;
+ uuid_str = NULL;
+
+ /* shd doesn't have a port. but the cli needs a port key with
+ * a zero value to parse.
+ * */
+
+ keylen = snprintf(key, sizeof(key), "brick%d.port", count);
+ ret = dict_set_int32n(dict, key, keylen, 0);
+ if (ret)
+ goto out;
+
+ pidfile = volinfo->shd.svc.proc.pidfile;
+
+ brick_online = gf_is_service_running(pidfile, &pid);
+
+ /* If shd is not running, then don't print the pid */
+ if (!brick_online)
+ pid = -1;
+ keylen = snprintf(key, sizeof(key), "brick%d.pid", count);
+ ret = dict_set_int32n(dict, key, keylen, pid);
+ if (ret)
+ goto out;
+
+ keylen = snprintf(key, sizeof(key), "brick%d.status", count);
+ ret = dict_set_int32n(dict, key, keylen, brick_online);
+
+out:
+ if (uuid_str)
+ GF_FREE(uuid_str);
+ if (ret)
+ gf_msg(this ? this->name : "glusterd", GF_LOG_ERROR, 0,
+ GD_MSG_DICT_SET_FAILED,
+ "Returning %d. adding values to dict failed", ret);
+
+ return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.h b/xlators/mgmt/glusterd/src/glusterd-utils.h
index 9bf19a6..3647c34 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.h
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.h
@@ -876,4 +876,8 @@ glusterd_is_profile_on(glusterd_volinfo_t *volinfo);
char *
search_brick_path_from_proc(pid_t brick_pid, char *brickpath);
+
+int32_t
+glusterd_add_shd_to_dict(glusterd_volinfo_t *volinfo, dict_t *dict,
+ int32_t count);
#endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-volgen.c b/xlators/mgmt/glusterd/src/glusterd-volgen.c
index 1f53beb..324ec2f 100644
--- a/xlators/mgmt/glusterd/src/glusterd-volgen.c
+++ b/xlators/mgmt/glusterd/src/glusterd-volgen.c
@@ -36,6 +36,7 @@
#include "glusterd-svc-mgmt.h"
#include "glusterd-svc-helper.h"
#include "glusterd-snapd-svc-helper.h"
+#include "glusterd-shd-svc-helper.h"
#include "glusterd-gfproxyd-svc-helper.h"
struct gd_validate_reconf_opts {
@@ -4845,7 +4846,7 @@ volgen_get_shd_key(int type)
static int
volgen_set_shd_key_enable(dict_t *set_dict, const int type)
{
- int ret = -1;
+ int ret = 0;
switch (type) {
case GF_CLUSTER_TYPE_REPLICATE:
@@ -5136,24 +5137,15 @@ out:
static int
build_shd_volume_graph(xlator_t *this, volgen_graph_t *graph,
glusterd_volinfo_t *volinfo, dict_t *mod_dict,
- dict_t *set_dict, gf_boolean_t graph_check,
- gf_boolean_t *valid_config)
+ dict_t *set_dict, gf_boolean_t graph_check)
{
volgen_graph_t cgraph = {0};
int ret = 0;
int clusters = -1;
- if (!graph_check && (volinfo->status != GLUSTERD_STATUS_STARTED))
- goto out;
-
if (!glusterd_is_shd_compatible_volume(volinfo))
goto out;
- /* Shd graph is valid only when there is at least one
- * replica/disperse volume is present
- */
- *valid_config = _gf_true;
-
ret = prepare_shd_volume_options(volinfo, mod_dict, set_dict);
if (ret)
goto out;
@@ -5183,19 +5175,16 @@ out:
}
int
-build_shd_graph(volgen_graph_t *graph, dict_t *mod_dict)
+build_shd_graph(glusterd_volinfo_t *volinfo, volgen_graph_t *graph,
+ dict_t *mod_dict)
{
- glusterd_volinfo_t *voliter = NULL;
xlator_t *this = NULL;
- glusterd_conf_t *priv = NULL;
dict_t *set_dict = NULL;
int ret = 0;
- gf_boolean_t valid_config = _gf_false;
xlator_t *iostxl = NULL;
gf_boolean_t graph_check = _gf_false;
this = THIS;
- priv = this->private;
set_dict = dict_new();
if (!set_dict) {
@@ -5205,26 +5194,18 @@ build_shd_graph(volgen_graph_t *graph, dict_t *mod_dict)
if (mod_dict)
graph_check = dict_get_str_boolean(mod_dict, "graph-check", 0);
- iostxl = volgen_graph_add_as(graph, "debug/io-stats", "glustershd");
+ iostxl = volgen_graph_add_as(graph, "debug/io-stats", volinfo->volname);
if (!iostxl) {
ret = -1;
goto out;
}
- cds_list_for_each_entry(voliter, &priv->volumes, vol_list)
- {
- ret = build_shd_volume_graph(this, graph, voliter, mod_dict, set_dict,
- graph_check, &valid_config);
- ret = dict_reset(set_dict);
- if (ret)
- goto out;
- }
+ ret = build_shd_volume_graph(this, graph, volinfo, mod_dict, set_dict,
+ graph_check);
out:
if (set_dict)
dict_unref(set_dict);
- if (!valid_config)
- ret = -EINVAL;
return ret;
}
@@ -6541,6 +6522,10 @@ glusterd_create_volfiles(glusterd_volinfo_t *volinfo)
if (ret)
gf_log(this->name, GF_LOG_ERROR, "Could not generate gfproxy volfiles");
+ ret = glusterd_shdsvc_create_volfile(volinfo);
+ if (ret)
+ gf_log(this->name, GF_LOG_ERROR, "Could not generate shd volfiles");
+
dict_del_sizen(volinfo->dict, "skip-CLIOT");
out:
@@ -6621,7 +6606,7 @@ validate_shdopts(glusterd_volinfo_t *volinfo, dict_t *val_dict,
ret = dict_set_int32_sizen(val_dict, "graph-check", 1);
if (ret)
goto out;
- ret = build_shd_graph(&graph, val_dict);
+ ret = build_shd_graph(volinfo, &graph, val_dict);
if (!ret)
ret = graph_reconf_validateopt(&graph.graph, op_errstr);
@@ -6998,3 +6983,22 @@ gd_is_boolean_option(char *key)
return _gf_false;
}
+
+int
+glusterd_shdsvc_generate_volfile(glusterd_volinfo_t *volinfo, char *filename,
+ dict_t *mode_dict)
+{
+ int ret = -1;
+ volgen_graph_t graph = {
+ 0,
+ };
+
+ graph.type = GF_SHD;
+ ret = build_shd_graph(volinfo, &graph, mode_dict);
+ if (!ret)
+ ret = volgen_write_volfile(&graph, filename);
+
+ volgen_graph_free(&graph);
+
+ return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-volgen.h b/xlators/mgmt/glusterd/src/glusterd-volgen.h
index f9fc068..897d8fa 100644
--- a/xlators/mgmt/glusterd/src/glusterd-volgen.h
+++ b/xlators/mgmt/glusterd/src/glusterd-volgen.h
@@ -66,6 +66,7 @@ typedef enum {
GF_REBALANCED = 1,
GF_QUOTAD,
GF_SNAPD,
+ GF_SHD,
} glusterd_graph_type_t;
struct volgen_graph {
@@ -77,6 +78,8 @@ typedef struct volgen_graph volgen_graph_t;
typedef int (*glusterd_graph_builder_t)(volgen_graph_t *graph,
dict_t *mod_dict);
+typedef int (*glusterd_vol_graph_builder_t)(glusterd_volinfo_t *,
+ char *filename, dict_t *mod_dict);
#define COMPLETE_OPTION(key, completion, ret) \
do { \
@@ -201,7 +204,8 @@ void
glusterd_get_shd_filepath(char *filename);
int
-build_shd_graph(volgen_graph_t *graph, dict_t *mod_dict);
+build_shd_graph(glusterd_volinfo_t *volinfo, volgen_graph_t *graph,
+ dict_t *mod_dict);
int
build_nfs_graph(volgen_graph_t *graph, dict_t *mod_dict);
@@ -313,4 +317,9 @@ glusterd_generate_gfproxyd_volfile(glusterd_volinfo_t *volinfo);
int
glusterd_build_gfproxyd_volfile(glusterd_volinfo_t *volinfo, char *filename);
+
+int
+glusterd_shdsvc_generate_volfile(glusterd_volinfo_t *volinfo, char *filename,
+ dict_t *mode_dict);
+
#endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
index 1ea8ba6..4c3ad50 100644
--- a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
+++ b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
@@ -1940,7 +1940,7 @@ static int
glusterd_handle_heal_cmd(xlator_t *this, glusterd_volinfo_t *volinfo,
dict_t *dict, char **op_errstr)
{
- glusterd_conf_t *priv = NULL;
+ glusterd_svc_t *svc = NULL;
gf_xl_afr_op_t heal_op = GF_SHD_OP_INVALID;
int ret = 0;
char msg[2408] = {
@@ -1950,7 +1950,6 @@ glusterd_handle_heal_cmd(xlator_t *this, glusterd_volinfo_t *volinfo,
"Self-heal daemon is not running. "
"Check self-heal daemon log file.";
- priv = this->private;
ret = dict_get_int32n(dict, "heal-op", SLEN("heal-op"),
(int32_t *)&heal_op);
if (ret) {
@@ -1959,6 +1958,7 @@ glusterd_handle_heal_cmd(xlator_t *this, glusterd_volinfo_t *volinfo,
goto out;
}
+ svc = &(volinfo->shd.svc);
switch (heal_op) {
case GF_SHD_OP_INVALID:
case GF_SHD_OP_HEAL_ENABLE: /* This op should be handled in volume-set*/
@@ -1988,7 +1988,7 @@ glusterd_handle_heal_cmd(xlator_t *this, glusterd_volinfo_t *volinfo,
goto out;
}
- if (!priv->shd_svc.online) {
+ if (!svc->online) {
ret = -1;
*op_errstr = gf_strdup(offline_msg);
goto out;
@@ -2009,7 +2009,7 @@ glusterd_handle_heal_cmd(xlator_t *this, glusterd_volinfo_t *volinfo,
goto out;
}
- if (!priv->shd_svc.online) {
+ if (!svc->online) {
ret = -1;
*op_errstr = gf_strdup(offline_msg);
goto out;
diff --git a/xlators/mgmt/glusterd/src/glusterd.c b/xlators/mgmt/glusterd/src/glusterd.c
index ff5af42..89afb9c 100644
--- a/xlators/mgmt/glusterd/src/glusterd.c
+++ b/xlators/mgmt/glusterd/src/glusterd.c
@@ -1533,14 +1533,6 @@ init(xlator_t *this)
exit(1);
}
- ret = glusterd_init_var_run_dirs(this, rundir, GLUSTERD_GLUSTERSHD_RUN_DIR);
- if (ret) {
- gf_msg(this->name, GF_LOG_CRITICAL, 0, GD_MSG_CREATE_DIR_FAILED,
- "Unable to create "
- "glustershd running directory");
- exit(1);
- }
-
ret = glusterd_init_var_run_dirs(this, rundir, GLUSTERD_NFS_RUN_DIR);
if (ret) {
gf_msg(this->name, GF_LOG_CRITICAL, 0, GD_MSG_CREATE_DIR_FAILED,
@@ -1815,6 +1807,9 @@ init(xlator_t *this)
CDS_INIT_LIST_HEAD(&conf->snapshots);
CDS_INIT_LIST_HEAD(&conf->missed_snaps_list);
CDS_INIT_LIST_HEAD(&conf->brick_procs);
+ CDS_INIT_LIST_HEAD(&conf->shd_procs);
+ pthread_mutex_init(&conf->attach_lock, NULL);
+ pthread_mutex_init(&conf->volume_lock, NULL);
pthread_mutex_init(&conf->mutex, NULL);
conf->rpc = rpc;
@@ -1895,7 +1890,6 @@ init(xlator_t *this)
glusterd_mgmt_v3_lock_timer_init();
glusterd_txn_opinfo_dict_init();
- glusterd_shdsvc_build(&conf->shd_svc);
glusterd_nfssvc_build(&conf->nfs_svc);
glusterd_quotadsvc_build(&conf->quotad_svc);
glusterd_bitdsvc_build(&conf->bitd_svc);
diff --git a/xlators/mgmt/glusterd/src/glusterd.h b/xlators/mgmt/glusterd/src/glusterd.h
index e858ce4..0ac6e63 100644
--- a/xlators/mgmt/glusterd/src/glusterd.h
+++ b/xlators/mgmt/glusterd/src/glusterd.h
@@ -28,6 +28,7 @@
#include "glusterd-sm.h"
#include "glusterd-snapd-svc.h"
#include "glusterd-tierd-svc.h"
+#include "glusterd-shd-svc.h"
#include "glusterd-bitd-svc.h"
#include "glusterd1-xdr.h"
#include "protocol-common.h"
@@ -170,7 +171,6 @@ typedef struct {
char workdir[VALID_GLUSTERD_PATHMAX];
char rundir[VALID_GLUSTERD_PATHMAX];
rpcsvc_t *rpc;
- glusterd_svc_t shd_svc;
glusterd_svc_t nfs_svc;
glusterd_svc_t bitd_svc;
glusterd_svc_t scrub_svc;
@@ -179,6 +179,7 @@ typedef struct {
struct cds_list_head volumes;
struct cds_list_head snapshots; /*List of snap volumes */
struct cds_list_head brick_procs; /* List of brick processes */
+ struct cds_list_head shd_procs; /* List of shd processes */
pthread_mutex_t xprt_lock;
struct list_head xprt_list;
pthread_mutex_t import_volumes;
@@ -219,6 +220,11 @@ typedef struct {
gf_atomic_t blockers;
uint32_t mgmt_v3_lock_timeout;
gf_boolean_t restart_bricks;
+ pthread_mutex_t attach_lock; /* Lock can be per process or a common one */
+ pthread_mutex_t volume_lock; /* We release the big_lock from lot of places
+ which might lead the modification of volinfo
+ list.
+ */
} glusterd_conf_t;
typedef enum gf_brick_status {
@@ -498,6 +504,7 @@ struct glusterd_volinfo_ {
glusterd_snapdsvc_t snapd;
glusterd_tierdsvc_t tierd;
+ glusterd_shdsvc_t shd;
glusterd_gfproxydsvc_t gfproxyd;
int32_t quota_xattr_version;
gf_boolean_t stage_deleted; /* volume has passed staging
@@ -624,7 +631,6 @@ typedef enum {
#define GLUSTERD_DEFAULT_SNAPS_BRICK_DIR "/gluster/snaps"
#define GLUSTERD_BITD_RUN_DIR "/bitd"
#define GLUSTERD_SCRUB_RUN_DIR "/scrub"
-#define GLUSTERD_GLUSTERSHD_RUN_DIR "/glustershd"
#define GLUSTERD_NFS_RUN_DIR "/nfs"
#define GLUSTERD_QUOTAD_RUN_DIR "/quotad"
#define GLUSTER_SHARED_STORAGE_BRICK_DIR GLUSTERD_DEFAULT_WORKDIR "/ss_brick"
@@ -680,6 +686,26 @@ typedef ssize_t (*gd_serialize_t)(struct iovec outmsg, void *args);
} \
} while (0)
+#define GLUSTERD_GET_SHD_RUNDIR(path, volinfo, priv) \
+ do { \
+ int32_t _shd_dir_len; \
+ _shd_dir_len = snprintf(path, PATH_MAX, "%s/shd/%s", priv->rundir, \
+ volinfo->volname); \
+ if ((_shd_dir_len < 0) || (_shd_dir_len >= PATH_MAX)) { \
+ path[0] = 0; \
+ } \
+ } while (0)
+
+#define GLUSTERD_GET_SHD_PID_FILE(path, volinfo, priv) \
+ do { \
+ int32_t _shd_pid_len; \
+ _shd_pid_len = snprintf(path, PATH_MAX, "%s/shd/%s-shd.pid", \
+ priv->rundir, volinfo->volname); \
+ if ((_shd_pid_len < 0) || (_shd_pid_len >= PATH_MAX)) { \
+ path[0] = 0; \
+ } \
+ } while (0)
+
#define GLUSTERD_GET_VOLUME_PID_DIR(path, volinfo, priv) \
do { \
int32_t _vol_pid_len; \
diff --git a/xlators/protocol/client/src/client.c b/xlators/protocol/client/src/client.c
index 2d75714..19f5175 100644
--- a/xlators/protocol/client/src/client.c
+++ b/xlators/protocol/client/src/client.c
@@ -46,7 +46,6 @@ client_fini_complete(xlator_t *this)
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
clnt_conf_t *conf = this->private;
-
if (!conf->destroy)
return 0;
@@ -69,6 +68,11 @@ client_notify_dispatch_uniq(xlator_t *this, int32_t event, void *data, ...)
return 0;
return client_notify_dispatch(this, event, data);
+
+ /* Please avoid any code that access xlator object here
+ * Because for a child down event, once we do the signal
+ * we will start cleanup.
+ */
}
int
@@ -105,6 +109,11 @@ client_notify_dispatch(xlator_t *this, int32_t event, void *data, ...)
}
pthread_mutex_unlock(&ctx->notify_lock);
+ /* Please avoid any code that access xlator object here
+ * Because for a child down event, once we do the signal
+ * we will start cleanup.
+ */
+
return ret;
}
@@ -2272,6 +2281,7 @@ client_rpc_notify(struct rpc_clnt *rpc, void *mydata, rpc_clnt_event_t event,
{
xlator_t *this = NULL;
clnt_conf_t *conf = NULL;
+ gf_boolean_t is_parent_down = _gf_false;
int ret = 0;
this = mydata;
@@ -2333,6 +2343,19 @@ client_rpc_notify(struct rpc_clnt *rpc, void *mydata, rpc_clnt_event_t event,
if (conf->portmap_err_logged)
conf->disconnect_err_logged = 1;
}
+ /*
+ * Once we complete the child down notification,
+ * There is a chance that the graph might get freed,
+ * So it is not safe to access any xlator contens
+ * So here we are checking whether the parent is down
+ * or not.
+ */
+ pthread_mutex_lock(&conf->lock);
+ {
+ is_parent_down = conf->parent_down;
+ }
+ pthread_mutex_unlock(&conf->lock);
+
/* If the CHILD_DOWN event goes to parent xlator
multiple times, the logic of parent xlator notify
may get screwed up.. (eg. CHILD_MODIFIED event in
@@ -2340,6 +2363,12 @@ client_rpc_notify(struct rpc_clnt *rpc, void *mydata, rpc_clnt_event_t event,
to parent are genuine */
ret = client_notify_dispatch_uniq(this, GF_EVENT_CHILD_DOWN,
NULL);
+ if (is_parent_down) {
+ /* If parent is down, then there should not be any
+ * operation after a child down.
+ */
+ goto out;
+ }
if (ret)
gf_msg(this->name, GF_LOG_INFO, 0,
PC_MSG_CHILD_DOWN_NOTIFY_FAILED,
--
1.8.3.1